The primacy of humans
An expert checks what is recorded in their name. A person decides whether an agent may use it.
About the research
Making digital intelligence second nature in regulated work.
For AI to help with regulated work, it needs access to the judgement of experienced people. Those people also need a way to check how it uses what they know. This is the problem behind Brightbeam’s applied research and its Continuous Intelligence Operations.
We build working tools alongside the papers. Publishing both lets others examine the reasoning, try the methods and tell us where they fall short.
01Purpose
Why this matters to Brightbeam
Brightbeam’s mission is to integrate digital intelligence into regulated enterprises so that it becomes second nature. That means understanding how people do their work and where they need help, then building AI into the systems and responsibilities already in place.
Our research concentrates on the decisions involved: what an agent needs to know, what people need to review and how to control a change once the agent is in use.
What we believe
Agent and human collaboration
Approvals, corrections and handovers follow the work in one shared record.
Audit and control
People review a proposed rule and its test results before it reaches the agent, and can withdraw it later.
02Research question
How should an agent use someone’s judgement?
We need to understand what the expert noticed and why it mattered, then decide when the same reasoning applies elsewhere. That requires evidence about the work and the people responsible for reviewing it.
Recording a decision does not establish that it was a good one. We need ways to assess the reasoning and recognise where it fails. A poor decision can be a useful test case without being suitable advice.
The same evidence shapes how AI is trained and tested. Practitioners help define what good performance looks like and which mistakes matter.
03People
Our research team.
Arsalan Shahid
LinkedIn: Arsalan Shahid (opens in a new tab)Gordon Suttie
LinkedIn: Gordon Suttie (opens in a new tab)Philip Black
LinkedIn: Philip Black (opens in a new tab)
04How we work
Research others can inspect.
Many of our essays begin in Digital Intelligence: State of the Art and Beyond, a regular Brightbeam forum. People building production systems bring technical developments and practical problems for discussion before we write about them.
- PaperResearch papers
Sets out the argument and the method.
Read the papers - CodeOpen-source projects
Runs the method, so others can test it.
Brightbeam on GitHub (opens in a new tab) - EssayEssays and talks
Works through the questions as they arise.
Read the writing
Work with us
Start with a task your team knows well.
Bring a task where experience affects the result, along with the people who do and review the work. We can choose a question to investigate together and agree what evidence would answer it.
We also work with universities and other research partners. If you are planning a proposal or a study on AI in regulated work, we would like to hear about it.