Ambidexterity in UN Agencies: Managing for Innovation through Autonomy

(1)

Leiden University – Faculty of Governance & Global Affairs

Master Thesis - Public Administration

Ambidexterity in

UN Agencies:

Managing for Innovation

through Autonomy

Student: Varun Kaul Supervisor: Dr. Christoph Johann Stettina Student ID: s1734970 Second Reader: Dr. Johan Christensen

(2)

1

“I hate the word innovation. It’s such a dangerous word being thrown

around. It means everything and nothing. But let me answer that with

slightly different phrasing. I think we should stop using the word

innovation, I think we should get rid of it. I think its legitimately a

dangerous word, because people start to believe that it’s a thing in

itself.”

(3)

2

Executive Summary

Dynamic crises increasing in scale, exponential growth of displaced populations, and the rise of fast-paced developing challenges, all were factors that characterized the humanitarian sector at the turn of the decade. Compelled to find new avenues to combat them, large humanitarian aid organizations like the United Nations (UN) recently turned to more fluid and adaptive measures for action. One such measure was the creation of ‘innovation units’. By 2016, though varied in form, innovation exists within 11 of the 15 UN agencies.

This thesis sets out to understand how the United Nations manages its various innovation units; with particular focus on the relationship between parent agency and innovation unit. Similar mapping and theorizing has been done successfully in the corporate sector. Studies on private sector innovation are now commonplace. This serves as a positive impetus to this thesis.

A multiple case study is done, with four UN agencies being selected. Data triangulation is employed, as part of which eleven semi-structured interviews, ten online surveys and document analysis of an extensive formative evaluation have been conducted. The data enables the answering of the central research question: How do UN parent agencies manage autonomy afforded to their innovation units? and sub-research question: What are the dimensions reflecting the autonomy that a UN parent agency grants to innovation units?

This study finds that UN innovation units are semi-autonomous and the dimensions of autonomy need to be managed as such. Despite a siloed approach, UN agencies are showing a shift towards four best practices with regards to managing their respective innovation units. Further, a structurally ambidextrous approach to better manage innovation within a large International Organization is proposed.

Given below are the practical recommendations that emerged from the thesis-

• Strategic direction should be provided by the parent agency and serve as a guiding tool for the innovation units.

• Financial/ revenue streams should be balanced between core and external funding. This funding will reflect a buy-in from the overall organization and maintain sufficient autonomy required to be innovative.

• Structurally, innovation should be firmly rooted within the overall organization. This includes a strong presence within headquarters.

• The model of autonomy proposed can be used by IOs to map and gain a preliminary understanding of their own innovation units. This model can and should be further developed. • Inter-agency communication should be further developed, as this would improve the sharing

and development of best-practices with regards to managing innovation within international organizations like the UN.

(4)

3

Acknowledgements

I want thank Dr. Christoph Johann Stettina for continuing to push this thesis beyond what I thought it could and should be. I want to thank him for acting as my supervisor, promotor and boss; providing me with comments that ripped apart this thesis, only to build it up. I would also like to thank my second reader, Dr. Johan Christensen for his comments, scrutiny and for burgeoning my interest in this topic through his lectures. Additionally, I would like to thank Thomas Baar for connecting me to my respondents and making sure that my requests don’t fall on deaf ears. More importantly, keeping me motivated, by reminding me of the potential implications of my thesis. Lastly, I would like to express my gratitude to my Jerome Gard and my anonymized respondents, who gave me their time.

I blame all of you. This is your fault. My overbearing family of six, who may not grasp what this thesis is about but I’ll be damned if they don’t support me through every second of everything I do. You say I’m a middle child of few words, well here’s over twenty-five thousand. I also want to blame Timna, because without you, quite literally this thesis would have been a pile of garbage. You are my entire support system, as I believe I am yours, so I guess that cancels out.

(5)

4

List of Tables

Table 1: Defining innovation in this thesis ... 9

Table 2: Innovation Typology (adapted Henderson & Clark, 1990) ... 16

Table 3: Question number to targeted dimension ... 32

Table 4: Operationalization of the code ... 34

Table 5: Case Overview- by the numbers ... 39

Table 6: Survey Responses Q2 ... 40

Table 12: Cross-case comparison (long) ... 44

Table 13: Interview Characteristics Case A ... 45

Table 14: Summarizing Case A ... 50

Table 15: Interview Charecteristics Case B ... 50

Table 16: Summarizing Case B ... 54

Table 17: Interview Charecteristic Case C ... 54

Table 18: Summarizing Case C ... 59

Table 19: Interview charecteristics Case D ... 59

Table 20: Summarizing Case D ... 63

Table 21: cf. Dette, 2016 ... 66

Table 22: Overview of comparison made... 67

(8)

7

List of Figures

Figure 1: Locating the proposed model of autonomy in research gap ... 23

Figure 2 The initial 10 dimensions of autonomy ... 24

Figure 3:Research Onion (cf. Saunders, Lewis & Thornhill, 2008) ... 30

Figure 4: Example question from Survey (see Appendix B for Survey Guide) ... 33

Figure 5: Structural layout of Case A ... 46

Figure 6: Innovation as a service provided by parent agency in Case A ... 49

Figure 7: Layout of Case B ... 53

Figure 8: Layout of Case C ... 56

Figure 9: Structural layout of Case D ... 60

Figure 10: Innovation matrix given by Henderson & Clark, 1990 ... 68

(9)

8

1. Introduction

This chapter introduces the research. First, a short background on the research is given, including defining terms that play a central role throughout this thesis. Next, the motivation for this thesis is given. Following which I present the central problem statement and research question that guides my thesis. Further, I provide the relevance of this research, both academic and practical. I conclude by giving an overview of the thesis.

1.1. Background

In 2017, the European Commission invested € 8.5 billion in innovation; 3% of a nation’s GDP directed towards Research & Development, and innovation is seen as the new norm by the European Union. United Nations (UN) investment in innovation amounts well over $10 billion with over 11 UN agencies and departments having well defined innovation cultures, which include, but are not limited to innovation teams, units, labs, funds or specialists (UNFPA, 2017; EC, 2018).

Having long been on the fringes of the UN mandate, innovation was identified in 2015 as one the organization’s seventeen sustainable development goals (SDG Nr.9). Innovation has since become a central resolution for the organization to achieve its core goals; and is increasingly forms an integral part of the United Nation’s long-term agenda.

Innovation has been observed to play a role in each UN agency or fund in both implicit (i.e. integrated in practices, initiatives or projects) and explicit (i.e. creation of innovation units or labs) manners; to the extent that, when an agency has “established new organizational strategies, innovation often features prominently as a goal or value” (Dette, 2016, p.4). As an integrated form of international governance, innovation units have developed at a rapid rate. They did so, in part, to respond to fast growing challenges in the field of humanitarian aid (Bloom & Faulkner, 2015). The present research focuses on understanding the new form of ‘governance through innovation’ that has developed in international organizations. However, as research traditionally lags behind innovation, there currently is a lack of understanding of the phenomenon (Savoia & Copeland, 2011).

(10)

9

This is in stark contrast with the private sector, which continually benefits from a growing research base on the functioning and management of innovation units. In the private sector, innovative initiatives, such as corporate ventures have become a long-standing pillar of success (Gard, 2015). Innovation is now thought to be key in securing a business’ adaptability and, thus, longevity. It is particularly salient in the technology sector, in which companies such as Amazon and Google are paving the way for others (Savoia & Copeland, 2011; Fiset & Dostaler, 2017).

Innovation units, throughout this thesis, are defined as “physical or virtual spaces that enable

and support the creative problem solving of those who participate in the space” (Bloom &

Faulkner, 2016, p.1374). These units are not stand-alone and are part of a larger agency. This larger agency is referred to as a ‘UN parent agency’ throughout this research. A UN parent agency defined as a United Nations’ agency, organization or fund that has commissioned the

innovation unit.

Concept Definition

Innovation Unit(s) “Physical or virtual spaces that enable and support the creative problem solving of those who

participate in the space” (Bloom & Faulkner, 2016, p. 1374).

UN Parent Agency Defined as a United Nations agency, organization or fund that has commissioned the innovation

unit.

Table 1: Defining innovation in this thesis Different Forms of Innovation within the United Nations:

Innovation does not appear to be streamlined throughout the UN. Each agency’s innovation unit departs from the same core ideal of innovation, and then develops its individual identity (Bloom & Faulkner, 2016; UNFPA, 2017). Currently, UN innovation exists in three identifiable forms:

(1) Physical spaces are set up to have a direct on-ground impact, presence and engage with communities. These larger innovation units increase their presence through ‘innovation champions’ or ‘innovation labs’ (Bloom & Faulkner, 2016)

(2) Virtual Labs prioritize open access, feedback and dissemination of information, to assist humanitarian aid. These labs can also work in tandem with physical units. For instance, the UN’s High Commission for Refugees (UNHCR) uses its virtual labs as a primary point to gain feedback and keep its innovative initiatives up to date (UNFPA, 2017).

(11)

10

(3) The final form taken by innovation is less explicit, rather organic. It consists of prioritizing and encouraging innovation within the International Organization (IO), through projects and initiatives. The International Organization for Migration (IOM), for example, has introduced innovative techniques in its everyday working. Projects such as IOMX -an “innovative campaign to encourage safe migration and public action to stop exploitation and human trafficking” (www.iomx.org)- are a direct result such organic form of innovation (IOM X, 2018).

1.2. Research Motivation & Objective

This research is motivated by the increasingly important role that innovation is playing in humanitarian agencies. This role has led to humanitarian organizations like the United Nations to go down the route of rapidly establishing innovation units. As mentioned previously, 11 out of the 15 UN core agencies have official innovation units. The lack of academic understanding about just how these innovation units work is what specifically drives this research.

The overall research objective is to understand how innovation units are managed within international organizations, and inform best-practices gaps (i.e. where and how management can improve). To achieve this objective, the research focuses on the facets of autonomy possessed by innovation units within the United Nations. By extension, it sets out to contribute to current understanding of ambidextrous management of innovation units within international organizations.

1.3. Problem Statement and Research Question(s)

Literature that sought to map UN innovation units within their respective parent agencies, reveal that innovation units show a certain level of embeddedness within the parent agency (Betts, 2014; Dette, 2016). A key aspect of such embeddedness is that innovation units are relatively autonomous. However, knowledge on how innovation units’ dual nature is managed -enabling them to be both embedded in and autonomous from a parent agency- appears insufficient. To address this lack of understanding, the following central research question is formulated:

(12)

11

Answering this question may help fill a significant gap in the study of ‘governance of innovation’, as well as provide a theoretical/analytical structure for further research in the field.

An integral assumption to this thesis is that UN innovation units possess a certain level of autonomy. This assumption is supported by empirical research conducted by Betts (2014, 2015), Bloom & Faulkner (2015, 2016) and Dette (2016). However, the exact nature of this autonomy is unknown; and the various facets that characterize it have yet to be theorized. Therefore, the problem statement is complemented with a more focused sub- question:

What are the dimensions reflecting the autonomy that a UN parent agency grants to innovation units?

1.4. Research Relevance

Theoretical Relevance: As noted by Schuurman & Tõnurist (2017), literature on UN innovation labs is sparse and “remains descriptive in nature” (p.10). The gap in scientific knowledge is significant. This thesis thus aims to move away from description, and towards understanding the workings of UN innovation. In building the first and only theoretical model of autonomy tailored specifically for the international organizations context, the thesis is expected to constitute an initial step in the relatively under studied topic of UN innovation management. In doing so, it also makes relevant contributions to the following research fields:

(1) Governance in Public Administration

(2) Private Sector innovation management research

Practical Relevance: Prior studies have been limited to cursory mapping and understanding motivations in innovation unitss. At the time of writing, no scientific or practitioner’s publication has gone in depth in studying the management of UN innovation units. There is a general lack of knowledge and communication around best practices, common challenges, and the future of UN innovation. The present research is capable of bringing significant insight into the three.

First, the eleven interviews conducted with four different UN agencies and their units provide primary insight on (i) how the units function; (ii) how both spectrum of the hierarchy (parent

(13)

12

agency and innovation unit) perceive their working relations, and (iii) how both spectrum of the hierarchy perceive the autonomy afforded to innovation units.

Second, the proposed theoretical model of autonomy allows outlining of several management needs and strategies. It can serve as a basis to develop a standardized tool to continuously evaluate an ever-evolving UN innovation culture.

Lastly, an initial comparison with innovation in the private sector (cf. Gard, 2015) sheds light on commonalities and differences between the two sectors. This can encourage policy-makers to find solutions tailored for the international organizations context, rather than resorting to plain mimicry of private-sector strategies.

1.5. Thesis Overview

This thesis contains six chapters, a brief description of each is given below.

Chapter 1: Introduction to Innovation in International Organizations

The first chapter contained a short background to the United Nations and innovation. Further, it briefly discusses the motivation behind the research, the research objective the problem statement and research questions and the overall relevance. It concludes with this thesis overview.

Chapter 2: Related Work and Theoretical Embedding

A literature review of the thesis is provided. The reader is given an understanding of UN innovation units; and the theoretical relationships that tie public sector innovation and management of private sector innovation. Additionally, a new five-dimensional construct of autonomy is proposed. Further, the chapter locates the thesis in the larger field of academia.

Chapter 3: Research Method & Design

The overall research method and design is laid out in this chapter. The research strategy is developed. The research design is then presented, which includes: data Collection, coding and case selection. Finally, measures taken to ensure validity and reliability are discussed.

Chapter 4: Results

The results of the data collection are laid out, with a cross case comparison of the initial results and a detailed description of each individual case. Each case is analyzed through

(14)

theoretical-13

thematic analysis, where the autonomy construct serves as a pre-existing code. The chapter concludes with a summary of the results.

Chapter 5: Discussion and Analysis

The results are then discussed in light of theory and insight from the interviews conducted. It discusses the three major findings of the thesis and concludes with a model for structural ambidexterity in the UN.

Chapter 6: Conclusion

A summary of the main findings and concluding thoughts are given in this chapter. I then give practical recommendations based on these findings. Finally, I highlight overall limitations and discuss further avenues for research.

(15)

14

2. Related Work & Theoretical Embedding

This chapter introduces the relevant literature and embeds the research in theory. First, I introduce the reader to innovation and innovation units within the United nations. Second, I locate this research in the field of public administration, by looking at concepts like governance innovation. Third, I locate this research in the field of management, by looking at management through structural ambidexterity. Finally, I discuss the research gap, challenges to UN innovation and present a theoretical model that will be used during this thesis.

2.1. Innovation Units & the United Nations

This section introduces the reader to innovation within the United Nations. First, the transfer of ‘innovation’ as an idea, from private sector companies to international organizations is discussed. Second, innovation in the United Nations is introduced. Finally, attempts to streamline innovation in the UN are discussed.

2.1.1. Innovation: From the private sector to the public

The private sector has regarded innovation as key to a thriving organization (Google, 2011; Teece et al., 1997). Conversely aversion to innovation is well documented in the public and humanitarian sector (Borins, 2002). Borins, puts this down to the traditionally large bureaucratic structures, that are international organizations. These international organizations are pre-structured to deliver on core objectives. In addition to this management of international organizations occurs in a ‘fishbowl’, i.e.- external actors are always waiting for the next failure in the public sector (Borins, 2002).

Whilst the private sector encourages fail-fast techniques to innovation, senior management in the public sector, have never been keen on committing to untested approaches. The rate of increase in established innovation units shows an increased investment in R&D and a shift away from the immediate response solutions are now taking place (Betts & Bloom, 2014). The overwhelming success of innovation cultures in private sector giants such as Google, Apple and Facebook, both at an individual and organizational level, was a major impetus in green lighting disruptive innovation units within IOs such as the UN. However international organizations need to move beyond this “private sector fetishization” (Simon, Nicholls &

(16)

15

Gabriel, 2015, p.33) as the challenges that face these two sectors are very different. As an example, overcoming the static UN bureaucratic structure has proved to be a central challenge to innovation (UNICEF, 2012); whereas private companies are now structured to allow for innovation (Fiset & Dostaler, 2017).

2.1.2. Innovation in the United Nations

The United Nations is comprised of a complicated set of interconnected agencies (Jolly, Emmerij & Weiss, 2009); and these agencies have their own siloed ways of inculcating innovation (Dette, 2016). At the time of writing this thesis, the United Nations does not possess a standard of innovation across its agencies, with each agency defining its own innovation culture. Innovation units need to be agile and quick to adapt; thus develop organically, regardless of which parent agency they may fall under. UN innovation labs vary in context, nature, structure and mobility (Bloom & Faulkner, 2015).

Louise Bloom, Romily Faulkner (2014, 2015, 2016) and Alexander Betts (2013, 2014) spearhead the research on UN innovation units. Their research accounting for a majority of the available material in current academic circulation.

Innovation in the UN is context specific and innovation units are uniquely tailored to the challenges of the agency that they are located in; as such, they change focus areas and goals according to their local context. This is a core ideal of UN innovation labs according to Betts (2014). To help them adapt structurally, most agencies have innovation labs in two forms-

physical and virtual.

UNICEF, UNHCR and UNDP are examples of agencies that are predominantly dependent on physical spaces, but have found a good balance between virtual and physical spaces (Dette, 2016); this results in them being quicker to adapt to changing situations. On the other hand, the International Organization for Migration (IOM, 2017) is an example of an agency that leans heavily on the virtual spaces, but is still in the initial stages of developing an innovation culture.

Purpose

The first imperative for innovation labs is an implicit one: organizational change. Bloom & Faulkner (2015) note that it is common to every innovation space within the UN. The imperative lies in organically promoting innovation through the organization. This is linked to

(17)

16

the belief that IOs can be more perceptive to disruptive change when it is inculcated from within, rather than injected externally (Bloom & Faulkner, 2016). UNICEF including innovation in its organizational strategy has been pointed out by Bloom and Faulkner (2015) as the perfect example of the first imperative.

The second purpose of UN innovation is more explicit: community enabling. Their role is to support, fund, fast track and help develop projects that originate from within the community from which location the unit operates. Innovation labs look to directly influence humanitarian aid and promote projects until the community has acquired well-developed measures to cope with humanitarian challenges (Bloom & Faulkner, 2016). Innovation units, in the long term, can thus be seen as a means to an end; or as Josh Harvey, the then head of the Kosovo Innovation lab, put it: “we want to work ourselves out of a job” (Harvey Interview, as cited in Bloom & Faulkner, 2016).

When looking at this purpose in the context of innovation typology proposed by Rebecca M. Henderson and Kim B. Clark (1990). It is important to note that it falls between incremental and architectural innovation. As it looks initiate an organizational acceptance of innovation (architectural innovation), but to do so organically (Incremental innovation) (Henderson & Clark, 1990).

Innovation Typology Definition (adapted from Henderson & Clark, 1990)

Incremental Innovation Incremental changes to existing services, to keep a business/ organization current and competitive.

Architectural Innovation Changing the way an organization works, whilst keeping the core values and ideals intact. Modular Innovation Changing a component within an organization, whilst leaving the overall structure intact. Radical Innovation Innovating to disrupt an existing / create an entirely new market.

Table 2: Innovation Typology (adapted Henderson & Clark, 1990) What do UN innovation units do?

A majority of UN innovation units focus on designing, prototyping and deploying tools that help reach development goals. An example of this is UNICEFs’ UReport. Which is a fast-track SMS system that engages the youth to send real time crises updates. Developed in Uganda, the product innovation was developed and deployed with the help of the community, in under a year (Dette, 2016). As of 2016, they have surpassed 3 million users across 34 countries (ureport.in, 2016).

(18)

17

Innovation units depend on capacity building through community engagement practices. These include training youth, employing individuals from within the community and undertaking empowerment and skills trainings; all these contribute to significant capacity building within the affected communities (Bloom & Faulkner, 2015).

Key partnerships, or what UNICEF (2012) calls “operational research” (p.21), are used to

enable innovation units to utilize all available resources in the community. Data coordination, funding and community outreach are all functions that benefit greatly through partnerships. To summarize, what broadly, the functions of innovation units are - developing products and services, engaging the community and creating sustainable partnerships to achieve the individual goals of the agency. These are broad functions, which vary on a, per agency or per innovation unit basis.

2.1.3. Continued attempts to streamline UN innovation

Conventional international agencies tend to exist within epistemic communities of similar functioning agencies (Haas, 1992); Innovation units are no exception. UN Global Pulse, which has an innovation unit primarily focused on data driven solutions and tracking in response to crises, has been pushing to streamline innovation, with the objective of achieving a more coherent understanding of the benefits that it provides to the humanitarian sector.

In 2011 Sara Farmer, chief architect of UN Global Pulse at the time, revealed that a pan UN bridge was being built that would enable network coordination across the UN system1_{; the}

blueprint for this was codenamed “Blue Hacks” (unglobalpulse.org, 2011). This streamlining was resisted by many agencies within the UN as if all agencies followed a set pattern of innovation, it would cease to be disruptive (Bloom & Faulkner, 2015)

Blue hacks did not materialize into an officially recognized network; however, the UN innovation network has been in operation since 2012. Now classified as an informal network, whose membership currently stands at 10 agencies, the UN innovation network is co-chaired by UNICEF and UNHCR; their primary objective remains furthering cross-innovation within

(19)

18

agencies. This effort has largely been unsuccessful. Currently, innovation within the UN is not streamlined (UNFPA, 2017; Bloom & Faulkner, 2015)

Innovation moves much faster than research, and as such there is minimal descriptive work on UN innovation. What the current literature does provide, are indications on how we may better understand these units. The next two sections construct the theoretical framework within which this research is situated. Section 2.2 embeds the research in the field of public administration. Section 2.3 then does the same in the field of management.

2.2. Theoretical Lens (Public Administration)

This section covers the first part of the theoretical lens within which this research is embedded. First, I introduce governance innovation and the specific sub-section of governance innovation that is relevant to this thesis. Second, I look at the United Nations as an expertise accumulator. Finally, I look at the relationship between expertise and autonomy within the context of the UN.

2.2.1. Governance innovation

Contemporary literature in public administration has laid the valuable groundwork that contributes to current understandings of innovation in the public sector (Edquist et al., 2001; Damanpour & Schneider, 2006; Walker, 2014; De Vries et al., 2015). Whilst the literature in the conventional field of public administration is not sufficient to provide a comprehensive theoretical framework, academic literature has laid down typologies, goals, antecedents and motivations of innovation within public administration. The types of innovation present in public administration have been collated by De Vries et al (2015). The heuristic model constructed in their work classifies four types of innovation that occur in public administration.

process innovation, which according to Walker (2014) entails direct improvement in the

working and efficiency of both internal and external processes. Edquist et al (2001) took this research further and divided the processes into administrative process innovation, which looks more at changes in the structure of decision-making and technological process innovation, where innovation occurs as a direct result of technological improvements. The second type of innovation, product innovation, occurs when products and/or services are made available to

(20)

19

the public. In most cases, this is done to stimulate an increase in efficiency and output (Damanpour and Schneider, 2009). Third, conceptual innovation can be seen when bureaucratic settings adopt a completely changed perspective on a specific problem that may help in providing a novel solution (De Vries et al, 2015).

Governance innovation is the final typology that has been explored by public administration

scholars, and is the most relevant for this research. This innovation occurs above the organizational level. This means it is not isolated to one unit, but rather an overarching disruption on how activities are financed, decision-making takes place and performance evaluation that goes on within a governance network.

Governance innovation is unique in that it favors creating network based systems that are less constrained by the traditional governance architecture. An initiative of innovative governance will tap new sources of financing, through partnerships with the private sector. Here, ‘new sources’ refers to resources that were “previously on the side-line and not fully exploited “(Moore & Hartley, 2008, p.16). Governance innovation has two major precursors/ antecedents to innovation in public administration.

It is influenced at the environmental level (De Vries et al., 2015). The environment here refers to external context within which the international organization operates -i.e. media salience, political relevance and isomorphism showcased by similar agencies- (Di Maggio & Powell, 1991). The second antecedent is the organizational level, which influences governance innovation particularly regarding organizational resources, capacity, and the amount of ‘slack’ or internal freedom to restructure, that the organization possesses. Autonomy, risk-aversion and leadership all account for larger organizational antecedents. (Walker, 2006; De Vries et al, 2015).

In their research for developing a “heuristic framework of public sector innovation” (p.17), De Vries et al., (2015) found that in fifty-five percent of the cases, governance innovation was directly influenced by the environmental level, and in thirty-nine percent of cases by the organizational level (n=181). A focus on the organizational level of governance innovation plays a central theoretical role in this thesis.

(21)

20 2.2.2. UN as an expertise accumulator

Peter Haas (1992) identified a gap between the rapidly growing scope of the United Nations, and the expertise that it was equipped with. He matched this gap with a subsequent growing trend of expertise accumulation. From 1950 to 1985, a budgetary shift away from ‘traditional concerns’ to more specialized agencies occurred. This trend occurred both at the individual and organizational levels. Boswell (2008) highlighted the development of a ‘third-UN’ as a continuation of this trend. Since 2007, in the field of humanitarian aid, the United Nations continued this “administrative empowerment of specialized knowledge groups” (Haas, 1992, p.85) through the development of its innovation units through multiple agencies (Jolly, Emmerij & Weiss, 2009).

2.2.3. Relationship between expertise & autonomy

The relationship between expertise and autonomy is a central relationship explored in management of public administrations (Weber, 1947; Ellinas & Suleiman, 2012; Bauer & Ege, 2016). The root of this relationship can be traced to when an international organization lacks the organizational capacity to combat to newly developing challenges. It resorts to adding units, teams or individuals that are equipped to deal with dynamic challenges. A direct relationship between expertise and autonomy has been identified, as when these expert arms have been added, they need to successfully managed (Ellinas & Suleiman, 2012). Autonomy in Public administration literature since Weber (1946, 1947) has been synonymous with control that the central actor can exert over its expert arms. The autonomy side of this relationship will be examined in the literature review.

2.3.

Theoretical Lens (Management)

This section covers the second part of the theoretical lens within which this research is embedded. First, I introduce the concept of ambidexterity and ambidextrous organizations; and locate this research within the paradigm of structural ambidexterity. Second, I discuss how ambidexterity is an effective tool in management. Finally, I show how ambidexterity has been used to manage autonomy effectively in the private sector.

(22)

21 2.3.1. Ambidextrous Organizations

Innovative units or teams in an international organization, just like any other organizational unit, must be managed (Teece et al., 1997). Due to the nature of innovation, the management of these units moves away from traditional management practices.

Not all innovative initiatives are instantly successful; innovation management becomes a key factor in determining how successful and well integrated an innovative practice, product or unit becomes (O’Reilly & Tushmann, 2011). One such successful tool of management is

AMBIDEXTERITY,which is defined in this thesis as the ability of an organization, public or private,

to manage two different units that have distinct goals.

An organization is ambidextrous when it has one arm working in its regular business model and one arm working on adaptive / new models; both arms work towards different immediate goals but broad common goals. Longevity and success of an ambidextrous organization is directly related to the successful management of these two arms. Management literature further splits organizational ambidexterity into two- structural ambidexterity & contextual

ambidexterity. (O’Reilly & Tushman, 2011; Fiset & Dostaler, 2017)

Contextual Ambidexterity, - “companies where any individual can be ambidextrous” (Fiset &

Dostaler, 2017); can be identified through high levels of individual initiative within an organization. This ranges from a high-ranking individual who is adept at multi-tasking, to an individual who explores and initiates innovation and disruption beyond their given job description (Birkinshaw & Gibson, 2004).

Comparatively, structural ambidexterity is better defined, both in literature and in practice, through the private sector. In practice, there is clarity to a structurally ambidextrous organization, as firms will have distinct units or groups, where some units will follow the traditional (exploitative) ways of doing things, and the remaining units will focus completely on new avenues (explorative). The long-term goals of the overall organization remain intact. My research is located in the perspective of structural ambidexterity.

(23)

22

2.3.2. Ambidexterity as an effective management tool

Jerome Gard (2015), analyzed the ambidexterity of the German IT industry by focusing on the management of corporate ventures. Corporate ventures are teams that exist within corporations which work towards developing new businesses, separate from the traditional focus of the company. These teams are commissioned to make a company more adaptable to the dynamic market. They have an entrepreneurial setup and can function as either (a) internal teams that are integrated with other teams in the company, or (b) “external sub-units” (Gard, 2015, p.2), that operate as independent entities.

Gard is quick to point out, that an effective mid-way, from a management standpoint, would be to have semi-autonomous corporate ventures. Giving these units freedom from the larger corporation “protects the corporate ventures from the managerial cognitions, inertia and short-term pressure of the mainstream business” (Gard, 2015, p.17). Still maintaining enough control to ensure that the corporate venture team is part of the larger company.

These corporate venture teams, similar to innovation units, are set up to encourage innovation within an orthodox structure. The results from Gard’s state of the art research that exploring the dimensions of autonomy provides incredible insight into the management of ambidextrous organizations. Gard’s (2015) research is an invaluable reference point, that gives this research insight into how ambidexterity is an effective way to manage for innovation.

2.4. A New Construct for Autonomy is Proposed

In this section, I propose a new five-dimensional theoretical construct for autonomy.

Models for autonomy are extensive in public administration and management respectively. Literature in public administration has given us multiple models to measure/classify autonomy in bureaucracies, public sector or international organizations (Christensen, 1999, 2001; Smullen, 2002; Verhoest, 2002; Verhoest, Peters, Bouckaert and Verschuere, 2004; Verschuere, 2007; Yesilkagit & Van Thiel, 2008). These models are contained to static structures of international and local governance that revolve around two primary kinds of autonomy: (a) the kinds of decision making capabilities an agency possesses; or (b) the lack of constraints that were put on the decisions made by the agency (Verhoest et al., 2004).

(24)

23

Similarly, the field of management contributes to current understandings of autonomy and ambidexterity (Breaugh, 1985; March, 1991; McGrath, 1995; Andersen, 2004; O’Reilly & Tushman, 2011; Gard, 2015), through models that extensively document the private sector. The models provided here focus on management, private sector decision making and most importantly, autonomy as a part of an ambidextrous company.

However, when looking at an innovative unit within an international organization, paradigm of analysis lands squarely in the middle of these two fields of literature. Current models cannot satisfactorily measure or further understanding of UN innovation units. Thus, a new theoretical construct is conceptualized below.

Figure 1: Locating the proposed model of autonomy in research gap

A literature review is done to discern the distinct types of autonomy that theoretically contribute to a better understanding of how autonomy is distributed in public sector innovation units. The dimensions of autonomy have been selected to reflect a balance between indicators of autonomy from within public administration and management literature (Bauer & Ege, 2016; Gard, 2015; Nokkola & Bacevic, 2014; Verhoest et al., 2004); each dimension contains a definition that has been adapted to UN innovation units. The final list will be summarized and serve as the pre-existing code for this study’s data analysis.

(25)

24 Five-Dimensional Autonomy Construct

A total number of ten dimensions of autonomy were considered (Figure 2). These were narrowed down to a list of five, after an extensive literature review. The narrowing was made in light of the following observations:

(i) one dimension overlapped: financial autonomy, given by both Verhoest et al (2004) and Nokkola & Bacevic (2014);

(ii) three dimensions did not apply or were too specific: academic autonomy, staffing

autonomy (Nokkola & Bacevic, 2014) and decisional autonomy (Gard, 2015);

(iii) two dimensions were merged due to similar definitions: structural autonomy (Bauer & Ege, 2016) and organizational autonomy (Nokkola & Bacevic, 2014).

Figure 2 The initial 10 dimensions of autonomy

The remaining five dimensions are defined below.

STRUCTURAL AUTONOMY

Structural autonomy is one of the most comprehensive dimensions developed in the field of public administration, tapered specifically to international organizations. Variance in structural organization affects the capacity of autonomous behavior exercised. Bauer and Ege (2016), measure structural autonomy by focusing on embeddedness, administrative variance, statutory powers and resource base.

Bauer and Ege (2016) analyzed fifteen secretariats of International organizations, which included four agencies of the UN and the United Nations itself. Their results revealed a

Structural Autonomy Financial Autonomy Functional Autonomy Decisional Autonomy

Staffing Autonomy Job Autonomy Financial Autonomy Academic Autonomy Strategic Autonomy Orgazational Autonomy

(26)

25

significant variance in autonomy based on different structured organizations; and a clear distinction between “autonomy of will” (Bauer & Ege, 2016, p.1024), the capacity to develop preferences within a subunit and the “autonomy of action” (Bauer and Ege, 2016, p.1025), the ability to translate these preferences into action. Further, they showed that an external unit will not always mean high structural autonomy; conversely, an embedded unit need not represent low structural autonomy (Bauer & Ege, 2016).

Dimension 1: Structural Autonomy is the extent to which the embeddedness of an UN

innovation unit within the overall agency structure influences its position vis-a-vis the UN parent agency (cf. Bauer& Ege, 2016).

STRATEGIC AUTONOMY

Distributing decision-making powers on strategic matters may be disruptive for conventional governance models. However, the innovation units are placed in a dynamic environment and when decisional authority is distributed amongst lower level managers in a dynamic environment, a positive effect on organizational performance can be seen. When strategic authority is structured in a bottom-up manner (Mintzberg, 1972, 1978), this underpins “the emergence of autonomous responsive initiatives” (Andersen, 2005, p.1288), that in the private sector, has led to successful strategic directions taken.

Dimension 2: Strategic Autonomy Allowing Innovation unit leaders/managers to make decisions regarding the overall strategy or mandate, without constraint from the parent agency/ senior management (cf. Gard,2015).

FINANCIAL AUTONOMY

The management of resources, sources of funding and expanse of an agencies resource base, according to Verhoest et al. (2004), determine the extent of financial autonomy.

Dimension 3: Financial Autonomy-the extent to which the innovation unit is dependent upon parent agency funding or its own isolated funds for its financial resources and the extent to which it is responsible for its own losses (cf. Verhoest et al., 2004)

(27)

26 FUNCTIONAL AUTONOMY

Functional experts are departments within an organization that focus on specific areas of work- R&D, finance, human resources, sales etc. These focus areas or ‘functions’ depend on the broader focus of the organization itself. Functional autonomy refers to the extent that these functional teams cross between the parent agency and the innovation unit. Functional

Autonomy as a sub-concept cannot be transplanted from the private sector; as UN innovation

is not, and has resisted being streamlined. The exact functions in question, may change depending upon the agency in question

Previous empirical examples (Gard, 2015) have shown that innovative ventures, which operate separate from the main corporate organization, that had their own functional experts enjoyed more autonomy than those organizations that had cross-functional teams (e.g. a human resources team that is shared by both the central company and corporate venture team).

Dimension 4: Functional Autonomy- The amount of functional expertise that is available to the innovation unit internally; and how much of the expertise is required externally (cf. Gard, 2015).

JOB AUTONOMY

Job autonomy, conceptualized by Breaugh (1985), deals with autonomy in the workplace at

an individual level. Work mode decision making is divided into three dimensions: method i.e. - the freedom to choose the procedures used in the workplace; scheduling i.e. - the freedom to control the timing / sequencing of their work mode decisions; and finally, criteria, i.e. - having a say in the way performance evaluation is conducted (Gard, 2015). Breaugh (1985) developed a seven-point scale on these three dimensions; the central question being, what is the extent to which employees consider they can make these decisions without coercion. This seven-point scale was further developed by Gard (2015) and adapted to measure ambidexterity in the private sector. Gard’s’ final 6-point Likert scale is an up to date version of Breaugh’s (1985) initial measure, but still keeps the original criteria, i.e.- scheduling, methods and procedure of evaluation.

(28)

27

Dimension 5: Job Autonomy Ability of an innovation unit manager/ staff member to make work-mode decisions without constraint (cf. Gard,2015).

2.5. Research Gap

In this section, I elaborate on the research gap towards which my thesis contributes. First, I discuss the cursory classification of innovation units as ‘semi-autonomous’. Second, I highlight the two main challenges to fostering innovation within an international organization.

2.5.1. Semi-autonomous nature of innovation labs?

Bloom and Faulkner (2016), determined that innovation spaces operate distinctly from their programme offices. The results, projects and outcomes of these labs are “translated back into a language and format that can be injected back into program offices” (Bloom & Faulkner, 2016, p.1384). This is a deliberate attempt by the unit to remain flexible, but remains unexplored beyond primary assumptions.

The flexibility of innovation units remains limited as these labs are not isolated from the larger UN system and organizational constrains; UN Global pulse is an example often used, as it responds to yearly research agendas (UN Global Pulse, 2014). Their study (Bloom & Faulkner, 2016) generalizes that units within the UN are semi-autonomous in nature, as “a certain degree of autonomy is required for an innovation space to have real flexibility” (p.21), and deliver on its goals. These generalizations do not go into further detail, as to what this nature means.

2.5.2. Challenges to innovation in the United Nations

There are two main challenges to fostering innovation within the United Nations. First, international organizations are rigid and pre-structured to deliver on core objectives (Borins, 2002). Literature in public administration (Aleinikoff, 2014; Bauer & Ege, 2016) notes that a big part of innovation within the public domain relates to shifting the bureaucratic stasis hat has developed over time. Changing this requires top-down support (Jolly, Emmerij & Weiss, 2009).

Second, a risk averse senior management/executive director/ Headquarters, has long been the de-facto challenge to innovation in public administration, particularly when the

(29)

28

organization in question, deals with humanitarian challenges (Borins, 2002). The sensitive nature of the problems that humanitarian agencies encounter, play a major part in this risk aversion. Whilst the private sector encourages fail-fast techniques to innovation, senior management in the public sector, have never been keen on committing to untested approaches. Innovation labs show that increased investment in R&D and a shift away from the immediate response solutions are now taking place (Betts & Bloom, 2015).

2.6. Summary

This chapter introduces innovation in the UN- adoption from the private sector, purpose and goals. The theoretical framework is located in two fields of study- public administration and management. I situate this thesis in the organizational level of governance innovation; I then explain how the UN acts as an expertise accumulator, and how expertise and autonomy are closely related in the field of public administration.

In the field of management, the concept of ambidexterity or managing two units within the same organization plays a vital role. I locate this thesis within the ambit of structural ambidexterity and explain how ambidexterity is an effective management tool, when it comes to managing innovative teams. I then present my five-dimensional model of autonomy and define each dimension.

Finally, I highlight the research gap that this thesis aims to fill. The next chapter will discuss the overall methodology used in this study.

(30)

29

3. Research Method & Design

This following chapter describes the overall research methodology and design used in this thesis. The research design was developed in response to the problem statement How do UN

parent agencies manage autonomy afforded to their innovation units? and; the subsequent

research question What are the dimensions reflecting the autonomy that a UN parent agency

grants to innovation units?

The first section, discusses the research approach and strategy taken. The second section, expands upon the research design. This includes methods of data collection, coding and case selection. The third section discusses the validity of this research. Finally, the chapter concludes by discussing the measures taken to ensure reliability.

3.1. Research Strategy

The section below outlines the research strategy and methodology used. The research onion (Saunders, Lewis & Thornhill, 2016), helps guide the formation of this strategy. Moving from the outer layer i.e. - the philosophy behind the research, towards deeper layers such as time horizons and methods used.

The philosophy of this research is situated in the interpretivist school, as it looks to use empirical data to explore how innovation units are managed within a large organization; it additionally attempts to understand the relationship between innovation units and the larger organization. The core purpose of this study is to further academic understanding on the working and management of innovation units; and inform best practices in the field of humanitarian aid.

The research approach is primarily deductive in nature, but also uses the empirical findings to further theoretical understanding on the topic. This duopoly means that whilst primarily deductive, and inductive approach is employed (Yin, 2009).

This approach is paired with a Qualitative case study. The qualitative approach is the best fit as it gives us the ability to understand how parent agencies and innovation units perceive their autonomy, which is an imperative part of my research. In addition, the time constraint of a

(31)

30

master’s thesis does not allow for a statistically significant data set to be collected, and since there is no such data se present, quantitative research has been ruled out.

The study is cross-sectional in its time horizon, as the study looks at its cases during one point in time. This is selected as the research wants to inform current best practices, and as such, needs to look at all four cases at the time of writing (Lavrakas, 2008).

The research is exploratory in nature; it hopes to contribute to a better understanding of UN innovation units. It uses a multiple case design; as innovation varies from agency to agency, a single case design cannot capture this variation satisfactorily and could not inform best practices. A replication Design ensures that research has rigidity and helps in comparing the private and public sectors (Yin, 2011). In Figure 3, a representation of the onion adapted to fit this research is given.

Figure 3:Research Onion (cf. Saunders, Lewis & Thornhill, 2008)

To summarize- This research is primarily a deductive qualitative multiple case study, which is exploratory in nature.

(32)

31

3.2. Research Design

In the following section, the research design is laid out. The research design is preceded by an extensive literature review, which not only embeds the design in current academic literature, but also gives the theoretical model of autonomy, which is used as a reference point during multiple stages of the design.

First, the data collection method is chosen, followed by the coding used to analyze this data. Finally, the justification for the case selection is given; this includes the sampling and number of cases selected.

3.2.1. Data collection

A triangulated approach to data collection has been employed (Altbach, 1984). This entails conducted expert, semi-structured interviews, Likert Surveys and Document analysis. This approach, in light of the number of cases, is seen as the ideal way to improve on overall validity and reliability. For the research to be sound it is crucial to select the correct data collection and sampling techniques.

Expert, Semi-structured Interviews

The first data point in this triangulation utilizes 11 semi-structured expert interviews. The interviewees are anonymized. The structure of the interview follows closely the guidelines provided by Harvey (2011); the central questions in these interviews follows the pattern of the proposed five-dimensional model of autonomy, which ties it closely to this research’s theoretical framework. The flexibility that a semi-structured interview provides, enables interviewees to answer in their own words and the funnel into a more focused answer. Following Yin’s (2009) case study protocol, Appendix A contains a detailed interview guide, which includes both sets of questionnaires used in the interviews.

Likert Survey

The second data point in the triangulation utilize 10 close-ended surveys, with respondents answering the questions on a six-point scale. The close-ended survey is conducted via Qualtrics and sent to the interviewee in advance.

(33)

32

The survey has been adapted from Gard (cf. 2015), with minor changes in wording, to make it applicable to the present case. Similar to the interview guides, two different survey sets were created; Set A was sent to employees within the larger UN parent agency and Set B sent to the employees in the respective innovation units. The possible responses are on a 6-point scale. The survey questions have been included in the survey guide in Appendix B.

This is seen as an effective methodology, as the Likert questions enable us to measure the specific targeted dimensions of autonomy and supplement the open questions asked in the interview allowing the respondents to reflect and elaborate on their previous answers. The table below shows which dimensions were targeted via the survey questions. The detailed questionnaire is given in Appendix B.

Interview Dimension of autonomy targeted via question

Q1 Screener question- no dimension targeted Q2 Job Autonomy Q3 Strategic Autonomy Q4 Functional Autonomy Q5 Structural Autonomy Q6 Functional Autonomy Q7 Overall Autonomy

Table 3: Question number to targeted dimension

The changes to the survey are restricted to minor wording changes, i.e. - changing words like ‘corporate venture team’ to ‘innovation unit’. This has been done after consulting with the author of the original work, with an understanding that limitations would be acknowledged. Qualtrics has been chosen, as it is the go-to survey tool in the UN humanitarian sector, and thus provides a degree of familiarity that makes the interviewees more receptive. The Likert questions are adapted from Gard (2015), and serves as a tool for direct comparison with corporate ventures and by extension, the private sector. Given below is an example from the survey, as it would appear in the Qualtrics tool:

(34)

33

Figure 4: Example question from Survey (see Appendix B for Survey Guide) Document Analysis

The document analyzed is the ‘Formative evaluation of the UNFPA innovation initiative’ (2017). The evaluation contained three volumes and one comparative strudy It has been conducted by the United Nations Population Fund, and contains three volumes and a cross agency comparative study. This third data point published in June 2017, was not part of the initial research design.

This late publication presented a unique opportunity as it was released during the data collection and results phase. Due to the relevance of this formative evaluation and the reliability of the data, I incorporated this data source as it provides us with a valuable insight to the four cases chosen in this study.

The formative evaluation completes the triangulation of results from three different methods of data collection i.e. - Interviews, Surveys and Document analysis; this adds robustness to the analysis phase.

(35)

34 3.2.2. Coding

A theoretical-thematic analysis of the interview is done. It is defined as “A method for identifying, analyzing and reporting patterns within data.”(Braun & Clarke, 2006, p.79). This is a deductive, top-down variation of thematic analysis. By choosing a theoretical rather than an inductive analysis, I sacrifice a detailed overall description of the interviews. This variation is chosen to have a richer description of data that is focused on revealing more about this thesis’ central research question. (Boyatzis, 1998; Hayes, 1997).

The interview data analysis is conducted using the five-dimensional model of autonomy as a pre-existing code (Table 4). As this coding already focuses the analysis to a great extent, this results in a detailed description of the management of UN innovation units. The analysis is confined to the semantic level (i.e.- not looking beyond what is said), keeping in line with the rules of theoretical thematic analysis according to Boyatzis (1998). This coding is further strengthened by the focused Likert survey questions that have been asked.

Concept Definition Example(s) of indicators

used

Structural Autonomy The extent to which the embeddedness of an UN innovation unit within the overall agency structure influences its position vis-a-vis the UN parent agency (cf. Bauer& Ege, 2016).

Structural definition; proximity to HQ; defined location

Strategic Autonomy Allowing Innovation unit leaders/managers to make decisions regarding the overall strategy or mandate, without constraint from the parent agency/ senior management (cf. Gard, 2015).

Mandate alignment; ability to make strategic decisions; freedom from strategic oversight

Functional Autonomy The amount of functional expertise that is available to the innovation unit internally; and how much of the expertise is required externally (cf. Gard, 2015).

Defined functions; availability of in-house functional expertise; number of functions Financial Autonomy The extent to which the innovation unit is dependent upon

parent agency funding or its own isolated funds for its financial resources and the extent to which it is responsible for its own losses (cf. Verhoest et al., 2004)

Funding streams; financial dependence; external funding; salaries

Job Autonomy The ability of an innovation unit manager/ staff member to make work-mode decisions without constraint (cf. Gard, 2015).

method; scheduling; criteria

Table 4: Operationalization of the code

As the structure of the questions is intentionally hybrid, I do not disregard any new dimension given by the experts interviewed; as theoretical thematic analysis “can evolve through the coding process (which maps onto the inductive approach)” (Braun & Clarke, 2006, p.12).

(36)

35 3.2.3. Case selection

This research is a multiple case study, and is exploratory in nature. The sample size for this study is 4 cases with 11 interviewees; which is seen as adequate to answer the research questions, remain academically relevant and in addition, this ensures staying within the confines of the time and resource constrain of a master’s thesis. The cases are carefully selected and have embedded units of analysis that will contribute to the study. Each case has a specific context, i.e.- the reason it has been chosen (Yin, 2009).

Sampling

This study’s total population comprise all United Nations agencies, funds and programmes. The total population being well over thirty entities. The limitations of time and resources, means that this research cannot account for the entire population (Yin, 2009). I have used two non-random sampling strategies in this study:

➢ Purposive Sampling

This is a form of sampling that requires enough knowledge on relevant features and characteristics of the total population in question. The researcher conducted informal discussions with experts and accumulated sufficient knowledge before determining the four cases that would be included in the study (Mosley, 2013).

➢ Snowball Sampling

This method of sampling involves being referred to other potential interviewees. This helps in accumulating relevant respondents in a tight knit community. However, this may harm the overall representativeness. (Goodman, 1961)

Number of Cases & Units of Analysis

There are minimum of two interviews conducted per case. These cases are embedded (as opposed to holistic), i.e.- each case has embedded units of analysis (Yin, 2011).

The embedded units of analysis are the two different sets of anonymized interviews that will be conducted in each case. The first embedded unit is the set interviews that are conducted with experts who are in the active staff of the respective parent agency. The second embedded unit is the set of interviews that will be conducted with individuals directly under the innovation unit or team. These can be project leaders, lab managers or challenge owners within the innovation unit. There are four cases overall, and the total number of data points

(37)

36

come to 11 interviews and 10 surveys and one extensive document. An overview of the cases is given below (Yin, 2011).

3.3. Validity

This section addresses the validity of the research. Validity covers the overall soundness of the research, and discuses measures taken to address the studies construct, internal and external validity (Yin, 2009). Challenges, and measures taken to address these challenges are given below.

Construct Validity

Conceptual frameworks, indicators and definition all fall under this category. How robust is the construct, from which inferences are being made?

This thesis formulates its concepts and models, based on extensive literature reviews. The model of autonomy is rooted and has been defined after comparing multiple models across the fields of public administration and management. Further, the basis for this construct- the autonomy model given by Gard (2015), was cross checked with the author. This was done not through an informal interview, which helps in safeguarding the construct validity. The model itself, was used to shape the interview questionnaire, (E.g. - questions guided by structural autonomy or job autonomy) which keeps it within the ambit of the framework (Harvey, 2011). Additionally, the construct has been critically reviewed by the supervisor and an external expert in the field to limit ambiguity.

Internal Validity

The transparency and internal quality of the research comes into question here, i.e.- was the study conducted in the right way? (Yin, 2009). Whilst there Is no direct causation explicitly stated, there are assumptions and a best practices model presented at the end. All interviews were analyzed by conducting thorough theoretic al-thematic analysis (Boyatzis, 1998; Hayes, 1997). The Likert survey was locked using the Qualtrics tool, which ensured that there was no difference/ alterations in questions; which ensures uniformity and adds to the validity. Further, triangulation of the data, follows that all findings and analysis was done by comparing the interviews, with Likert surveys and documents.

Ambidexterity in UN Agencies: Managing for Innovation through Autonomy