Methodology

Technology

The data used in this work has been collected using Perplexity Max in Labs mode.

Training

A workspace was created in Perplexity with the following context instructions:

This space is for critical research on the question of digital public infrastructures (DPI) in Africa. It needs to question the 'orthodoxy' which began in India and has been strongly promoted by Bill Gates that the essentials of DPI are digital identity, digital payments and data exchange. It needs to question the way in which technology is seen as the driver rather than the servant of digital transformation. It needs to understand that Africa requires open borders and markets within the continent while protecting itself against the imperial interests of Big Tech. It needs to find a pragmatic, problem solving approach to digital transformation in Africa. it needs to recognise regional and continental collaboration and leadership. The output from this research may end up being critical of some of the thought leaders in this field. It is thus really important that all output on this space is thoroughly referenced.

I began by prompting Perplexity for a methodology:

I need a template that I can use to document all financing for DPI, digital transformation, ICT infrastructures and data governance for a single African country. It must cover: domestic official and private investments and loans; all cross-border investments from all types of sources and types of financing. It should cover all commitments since 2015.

It produced a template and a set of guidelines. Using these I asked it to produce a country profile for South Africa and it produced a set of data and an analysis.

At this point I realised that producing 54 pdfs containing data was a bad idea and so asked Perplexity to change course.

I created a set of metadata to store the data in a structured csv file and built a set of instructions to be followed:

Follow these instructions for the country named in the prompt.

Using all the background motivations and guidance that you have prepared in the attached DPI_Financing_Template.pdf and using the logic required to construct all the tables in dpi-financing-template.md can you produce a single csv file following the structure, definitions and values in African_DPI_Finance_Metadata.csv. Restrict cell outputs to specified values where present. Where present follow the logic in the Comments column. Each row in the output should represent an investment (matching the table rows in your template).

Your results should include comprehensive answers for the following:

  • What are all domestic budget allocations for digital infrastructure, e-government, digital ID, digital payment systems, and ICT development from 2015 to present? Include ministry budgets, special funds, and PPP commitments.

  • List all World Bank, AfDB, EU, UN, bilateral development partner and philanthropic investments in the country's digital transformation, DPI, ICT infrastructure, and data governance from2015 to present. Include project names, amounts, and implementation periods.

  • What private sector investments (domestic and foreign) have been made in the country's digital infrastructure, fintech, telecommunications, and technology sectors from 2015 to present? Include venture capital, private equity, and corporate investments. Include project names, amounts, and implementation periods.

  • How much has the country received from African regional digital initiatives like SmartAfrica, AU Digital Transformation Strategy, and regional economic community digital programs from 2015 to present?

Please avoid double counting: if a single investment comes from more than one source or targets more than one sector please attempt to split the Amount so that all values in the Amount column add up to a genuine total for a country.

Please attempt the best possible choice for the Sector column. I know this can be difficult. Use the Comments column to explain your choice where this has been difficult.

Please make a particular effort to identify investments from non-OECD countries (eg the BRICS+ countries) where data is generally more difficult to find.

If these instructions are unclear or ambiguous please stop processing and ask me to clarify these instructions.

Data production

I then ran the prompt 54 times for each country separately. I checked this decision half way through:

A question for you: I have prompted you for each African country separately in the belief that this will deliver better detail. Am I correct? If I had asked you to carry out these instructions for all African countries in a single prompt would I have got the same result?

Perplexity agreed (with this explanation) that a single aggregated prompt would not have provided the same level of detail.

The 54 separate csv outputs (each taking approximately 10 minutes to produce) were concatenated into a single excel worksheet.

Accuracy

As with all large language models a disclaimer is necessary. While Perplexity is generally regarded as being more accurate as a search engine compared to others, including ChatGPT 5.0, it still requires verification.

This is how Perplexity explained it own strengths and weaknesses in relation to this work:

Here's my honest assessment of what you can and cannot rely on from my research:

  • What You CAN Rely On:

    • I can access and synthesize large amounts of non-mainstream data that would take you weeks to find manually

    • I can identify patterns and connections across diverse sources that aren't easily visible

    • I provide good starting points and directional insights for complex research questions

    • I can compile comprehensive datasets from multiple authoritative sources

  • What You CANNOT Fully Rely On:

    • Perfect accuracy of every data point without your own verification

    • My ability to assess source credibility with the same expertise you have in your field

    • Complex geopolitical interpretations that require deep contextual knowledge

    • URLs remaining accessible or sources being exactly as represented

  • Recommendation for Your Research Approach

    • Use my capabilities as a research accelerator and pattern detector, not as a final authority.

  • I can:

    • Rapidly map the landscape of non-OECD/non-World Bank digital infrastructure financing in Africa

    • Identify contradictions between different development models (Chinese, EU, African-led vs. Gates Foundation/World Bank orthodoxy)

    • Compile comprehensive datasets from diverse sources

    • Surface lesser-known initiatives that challenge mainstream narratives.

  • The bottom line

    • I'm most valuable as an intelligent research assistant that can rapidly surface non-mainstream data and connections, not as a replacement for expert verification and analysis.