A Conversation with Raj Nair, Governance Advisor within the Commonwealth Secretariat

Edition #101 After the Commonwealth launched an AI-driven policy tool late last year, interweave finds out just what goes into developing a multilateral AI assistant...

A few months ago, the heads of state and senior political figures of the Commonwealth nations met for a five-day summit in Samoa. As usual for the biennial summit, the “class photo” was striking for its diversity, with fifty-six member states represented from every continent on the globe.

The Commonwealth CHOGM Summit in Samoa (Source)

As a multilateral body, creating AI tools to support such a range of nations is no mean feat. But back in August last year, the Commonwealth Secretariat nonetheless launched StrategusAI, an AI-driven policy toolkit developed in collaboration with Intel.

The LLM-driven tool is designed to support policymakers in developing national AI policies by providing references and granting them access to global expertise and best practice from entities such as the OECD and World Bank, allowing governments to create policy drafts with limited time and budget.

When I sit down with Raj Nair, the Governance Advisor within the Commonwealth Secretariat who is responsible for overseeing the development of this tool, he starts by putting this in the broader contexts of the Commonwealth’s efforts to support member states with their AI.

“This is not an isolated tool”, he tells me, “but part of a broader effort within the Commonwealth Artificial Intelligence Consortium (CAIC) formed off the back of the 2022 Commonwealth Heads of Government Meeting in Rwanda”.

There, “leaders reaffirmed their commitment to empowering citizens to harness the benefits of AI”, particularly in supporting the 33 small commonwealth states who lag in AI infrastructure.

Building a multilateral AI tool

When the CAIC was formed, “the number of Commonwealth countries without AI strategies was high – 65% of the Commonwealth did not have anything at all”. To develop a strategy, “they would need to have expensive consultants come in, and often by the time that they had a strategy ready, the space would have moved on”.

To tackle this challenge, CIAC was set up with four strands: policy development (led by Rwanda); Research & Innovation (led by The Gambia with Australia’s National AI Centre); Capacity Building (led by Bangladesh and supported by UC Berkeley); and Data & Infrastructure (led by Malta, with support from Nvidia).

Aside from tools like StrategusAI (and others including public debt management and governance performance measurement systems), Nair lists outputs from CIAC including an AI incubator to “make-a-thons” and efforts to upskill academia.

As we turn our attention to the AI assistant itself, Nair is keen to stress the plurality of actors involved in developing it. “Alongside the stewardship of the government of Rwanda, the tool uses input from Oxford Insights, the Sustainable Living lab and policymakers from other Commonwealth countries”.

All-in-all, in developing the tool, Nair’s team interviewed 17 countries, drawing on their collective intelligence to determine its inputs and use cases.

“Obviously these countries are very different”, he tells me, “but the processes for technology policymaking are very similar, and we realized that this process can be supported by AI automation”.

Developing AI policy with StrategusAI

We turn our attention to the functionality of the tool. Currently being piloted in Fiji and Brunei Darussalam, the tool is designed to be used by policymakers to quickly and efficiently generate AI policies.

With prompts directed a section at a time, the tool is able to create first draft individual sections of a policy one at a time, drawing on a base of similar country contexts and ambitions.

“People often ask how it is different from ChatGPT or other LLMs” Nair ventures. “It’s a combination of things – firstly, the process that you follow is similar to that of any other policy process, there is a structure in how the tool works. Secondly, you are given a knowledge base of vetted sources – such as the OECD – where you can trace back where the information is coming from.”

By automating elements of the policymaking process, such as research and early drafting stages, governments stand to gain significantly from using the tool. The Commonwealth estimates the costs of developing a government strategy as “ranging from US$250,000 to US$1.5 million in consulting fees, with development taking from six months up to a year, not including implementation phases”.

By using StrategusAI, Nair estimates, countries will be able to reduce this to a matter of weeks.

As with all AI tools, StrategusAI is not designed to provide a replacement to rigorous policymaking and user-centered design. “Traditionally, the policy development process requires a team of experts, but with the tool public sector leaders can tap into the global community from multiple entities”.

But the onus of developing priorities, choosing which other countries to align with, and innovating beyond peers still remains on governments themselves.

“There is still a delta between what a consultant could do and what AI could do”, Nair clarifies, but “it is able to do 50-60% of the work, and allow the governments to focus on delivering the policy”.

Getting a tool like this right could be a game changer for governments. In late 2023, The Times in the United Kingdom announced that Ministers there were trialing the use of AI tools for ministers to “save time, improve the policy-making process and reduce bureaucracy”.

The Deputy Prime Minister at the time, Oliver Dowden, explained the tool’s ability to bring together various submissions on a particular area of policy and recommend an array of outcomes.

But that particular example from the UK, itself a Commonwealth country, also highlights the delicacies over tools like this. Rupert McNeil, the former head of HR in the UK Civil service, estimated in June last year that the number of civil servants could be cut by 70% over the next decade thanks to Ai and automation (layoffs of 350,000 people).

The line between policy and contextual expertise, and the efficiencies delivered by AI tools, remains blurry.

In the Commonwealth too, potential job losses have been a key theme. Nair mentions that “we very recently had a Heads of Public Services Meeting” where he presented the CAIC, “and what came out very clearly is that there is a fear about what is happening to jobs”. Governments “both recognize that AI is the future, but worry about governance in multiple aspects”.

Solving that puzzle in a policymaking context, retaining expertise through human interaction and saving time from AI-tools, will be crucial to StrategusAI’s success, and the success of those strategies that it helps to produce.