The Landscape, in Data

What the work involves, sector-by-sector

82% of use cases are tagged Service Delivery or Data Analysis, the two most common classifications.

Service Delivery

294

Data Analysis

284

Access & Communication

194

Monitoring & Evaluation

185

Knowledge & Research

125

Infrastructure & Enablement

86

Content Creation

72

Operations & Project Management

44

UNIQUE USE CASES

Health & Nutrition

114
107
66
55
39
26
12
20

192

Education

101
38
40
28
20
13
30
8

125

Climate & Environment

15
85
20
84
30
27
8
6

116

Social Protection & Inclusion

37
16
49
8
9
8
13
4

68

Agriculture & Food Systems

25
36
21
19
5
10
8
1

47

Governance

9
21
14
16
19
6
6
3

43

Livelihoods & Economic Opportunity

30
15
20
3
4
6
2
3

39

Humanitarian & Disaster Response

12
27
10
22
5
3
1
4

35

Justice & Legal Systems

10
7
15
5
9
1
3
3

23

Domain-agnostic

1
5
2
1
6
8
2
1

12

CLICK ANY CELL TO VIEW THE USE CASES BEHIND IT
A USE CASE CAN STRETCH ACROSS MULTIPLE DOMAINS AND APPLICATIONS

Low

High

Which AI technologies are being used in each sector

In 9 of 10 sectors the leading AI technology is foundation models; in 2 it is classical ML and statistical methods.

Foundation models

394

Classical ML and statistical methods

277

Bespoke deep learning

189

Symbolic, optimisation and decision methods

50

UNIQUE USE CASES

Health & Nutrition

122

101

57

22

192

Education

109

47

29

9

125

Climate & Environment

53

81

53

11

116

Social Protection & Inclusion

59

21

27

3

68

Agriculture & Food Systems

30

29

22

3

47

Governance

36

25

18

2

43

Livelihoods & Economic Opportunity

33

16

9

4

39

Humanitarian & Disaster Response

21

21

18

3

35

Justice & Legal Systems

19

5

1

2

23

Domain-agnostic

9

1

2

2

12

Leading AI technology for each sector

Definitions

Foundation models

394

Very large pre-trained models directed by describing the task in words.

Classical ML and statistical methods

277

Statistical models fitted to rows-and-columns data.

Bespoke deep learning

189

Neural networks trained or adapted by the team for one narrow task.

Symbolic, optimisation and decision methods

50

Decisions from stated rules, objectives and constraints rather than learned patterns.

A USE CASE CAN RECORD MORE THAN ONE AI TECHNOLOGY CATEGORY

What is the scale of investment, by sector

45% of use cases need a custom build and a cross-functional team, while 6% can be generalist-led with no coding.

SECTOR
SHARE BY SCALE OF INVESTMENT
N

Humanitarian & Disaster Response

35

Governance

43

Justice & Legal Systems

23

Climate & Environment

116

Education

125

Social Protection & Inclusion

68

Domain-agnostic

12

Livelihoods & Economic Opportunity

39

Health & Nutrition

192

Agriculture & Food Systems

47

Definitions

High

254 · 45%

Requires a custom build and significant cross-functional internal team capacity (data scientists, engineers, program staff, ethics/compliance officers, etc.).

Medium

121 · 21%

Deploys turnkey solutions with some customization/fine tuning (at least 1 ML engineer, data scientist and back/front end development).

Low

33 · 6%

No coding required, and can be led by a generalist.

Not specified

161 · 28%

Not recorded for this use case.

How far along are AI use cases by sector

59% are deployed and running. The rest are still in the ideation or pilot phase.

Health & Nutrition

192

Education

125

Climate & Environment

116

Social Protection & Inclusion

68

Agriculture & Food Systems

47

Governance

43

Livelihoods & Economic Opportunity

39

Humanitarian & Disaster Response

35

Justice & Legal Systems

23

Domain-agnostic

12

Ideation

Pilot

Deployed

Discontinued

Where open resources are available

133 use cases are openly available; and the top 3 sectors account for 67%. Click any sector to break it down by AI technology.

SECTOR
OPENLY AVAILABLE
OF ALL

Open source

Source available

PROPRIETARY AND UNSTATED LICENCES NOT SHOWN