In 2022, I sold my AI company(Augmented Pixels)—an industry leader in autonomous navigation—to the international public company Qualcomm. Then, I launched a new business, 12New.ai, to disrupt sectors like compliance and finance with AI.

It turned out that our solutions were needed not only by corporations but also by international public institutions and anti-corruption organizations in various countries.

For example, at the beginning of 2024, one well-known anti-corruption organization, CHESNO, began using our solution to analyze draft legislation before it’s considered in parliament. Over the past two months, nonprofits in three more countries have started actively using our solution for similar tasks.

From my practical experience, AI provides substantial benefits for anti-corruption analysts: when reviewing proposed laws, they spend significantly less time on technical analysis and routine tasks—this work is done by AI, freeing people up for more creative tasks(like analyzing connections at a more general, abstract level, among other things).

Additionally, with the introduction of AI tools for analyzing draft laws, the government can no longer “play the classic game” of sending draft laws to activists only 1–2 days before a vote, leaving no time for thorough examination. Our AI tools allow analysts to produce a basic report within 1-2 minutes, effectively catching up to 90% of questionable provisions. We’ve confirmed these results with CHESNO and other nonprofits in four countries, and through this case, we see the potential for a qualitative shift in the fight against corruption through the adoption of AI.

Therefore, I believe that in the next 2–4 years, not only in the anti-corruption industry but also in the public sector as a whole, we can eliminate a large portion of inefficiency thanks to AI.

Where can we achieve the best results?

Right now, we’re involved in various efforts to improve the operations of large corporations and government institutions. We’re also analyzing bureaucratic systems and specific sectors that have historically been most prone to corruption.

Public Procurement

The first area where significant improvements can be achieved through AI tools is public procurement. According to Webinarcare, corruption leads to losing 10–25% of the total value of government contracts—approximately $950 billion. For comparison, according to the Food and Agriculture Organization of the United Nations (FAO), annual investments of about $267 billion are needed to achieve zero famine by 2030. In other words, the amount stolen annually in public procurement could eliminate global famine almost three times over (okay, slightly less than three times, but still).

In several countries, AI is already demonstrating significant success in combating corruption in procurement, as evidenced by data from the OECD (Organisation for Economic Co-operation and Development). Although the organization points out some risks—mainly related to the quality of the datasets used to train these AI-based solutions—I predict AI’s role in combating procurement corruption will only increase in the coming years.

Licensing and Permits

Another area that can be significantly improved with Generative AI is issuing government certificates and permits. I’ve been talking to government representatives from various countries in this field, and I see that decisions on most such permits—like retail licenses or private construction permits—can already be made autonomously by AI (and impartially, unlike “occasionally biased” officials).

Of course, at the start, you need a “human in the loop” in such processes, but overall, from our research, AI can help reduce the number of personnel involved in this process by 90%. That is, 90% of the work is done by AI, and 10% by people who review decisions or handle cases where users disagree with the AI’s decision.

Social payments

AI can find unusual patterns in payment requests, identify hidden connections, and identify discrepancies in the documents provided with the criteria for receiving social benefits.

It is also quite important to effectively implement post-analysis of people's behavior and see whether they actually use social benefits for their intended purpose, whether they have undeclared income and use social benefits or assistance for luxury goods, etc.

The introduction of AI into the processes of calculating social benefits would significantly complicate corruption in this area.

According to conservative estimates, automation could reduce the share of fraudulent transactions by at least 80%, and at the same time speed up the calculation of benefits by 2-3 times.

In Summary

In the long term, anti-corruption AI tools will become accessible not only to national-level nonprofits like CHESNO but also to activists at regional levels, significantly enhancing the quality and transparency of governance both locally and nationally.

AI will be used not only by activists to analyze the efficiency of officials but also by government agencies themselves to improve work quality. Smart integration of technology into business processes, combined with appropriate re-engineering where AI + HITL (human in the loop) replace old-fashioned government departments (where hundreds of officials shuffle papers around), will fundamentally change how the state operates. It will make the government dozens of times more flexible and efficient and much less susceptible to corruption.

I see that small countries, historically more open to change, are already practically implementing AI and achieving real results.

Larger states are often more conservative and cautious. They have massive legacy systems and, as a result, cannot quickly adapt—something that I believe does not work in their favor in the long run.

In my estimation, in the next 2–4 years, countries that do not implement the AI + HITL approach into their bureaucratic apparatus will gradually lose their competitive advantage on the international stage (including becoming operationally very inefficient).

Those who are already starting to build real systems, change their architectures, and train their people will come out on top.