What organisations found when they worked with us

We engaged Anggun Systems for the knowledge retrieval service. Our main concern going in was whether the system would actually work reliably enough for our compliance team to use in practice, or whether we would end up with something that looked impressive in a demo and failed in production. The evaluation report they delivered before deployment was the most reassuring part — they measured the citation accuracy on a sample of our actual documents and told us what the miss rate was. It was higher than I had hoped but lower than I had feared, and we knew exactly what we were dealing with before we deployed. The ninety-day tuning period genuinely improved the numbers.

April 2025 · Retrieval System engagement

The AI policy drafting engagement was useful but I would say to prospective clients: expect to do a meaningful amount of work in the review sessions yourself. Faris brought a draft that covered everything we had discussed, but there were sections specific to our sector's regulatory situation that required us to provide context they could not have known in advance. The final document was solid and has since been adopted by our board. My one suggestion would be that the initial interview could have been longer to capture more of this context upfront.

March 2025 · AI Policy Drafting engagement

We brought Anggun Systems in for the custom model engagement. What impressed my team technically was the care they took with the evaluation — they used a test set we had not seen them work with, and they flagged two categories of input where the model was notably weaker than on the overall test. That kind of honest reporting is not common in my experience. The handover documentation was thorough enough that we have since made two configuration changes ourselves without needing to contact them.

April 2025 · Custom Model Development

The retrieval system we built with Anggun Systems now handles queries against seven years of audit reports, internal guidance notes, and board minutes. The system cites its sources, which is non-negotiable for audit work — you cannot have staff relying on an AI response without being able to trace it back to the original document. This was a requirement they built in from the start rather than something we had to ask for twice.

March 2025 · Retrieval System engagement

We were a manufacturer looking at automating some of the classification work in our quality control process. The scoping conversation was direct — they told us what they could do, what they were uncertain about given our data, and what they would need to confirm before committing to a timeline. The model delivered was slightly below the accuracy threshold we had hoped for on one category, and they told us that before delivery rather than after we signed off. We are considering a follow-up engagement once we have collected more labelled data in that category.

April 2025 · Custom Model Development

Our firm had been using various AI tools informally for about eighteen months when we decided we needed a written policy. The document Faris drafted covered three things we had not specifically asked about but which were clearly relevant to our practice: how to handle client data when using AI tools, how to disclose AI involvement in work product, and how to keep the policy current as the tools change. These sections became the most discussed by our partners during review.

April 2025 · AI Policy Drafting engagement