Oops! Something went wrong while submitting the form.
Not a summary of what's on page one of Google. Sai works a market the way an analyst does — sizing it, mapping the players, finding the buyers' actual complaints, and telling you where the evidence runs out.
Market research used to be gated by data. It is now gated by synthesis. Everything you need for a first-pass view of most markets is public — analyst summaries, funding announcements, review sites, job postings, earnings commentary, forum threads where the actual buyers complain in specific detail. The work is not finding it. The work is reading enough of it to know which parts are load-bearing and which are one consultancy's projection being quoted back and forth until it sounds like consensus.
That is the job worth handing to an agent, and it is also where most AI research falls down. Ask a model to research a market and you typically get a fluent, confident document that reads well and cannot be checked. The fluency is the problem: an unsourced claim and a well-evidenced one look identical on the page.
The difference between output you can put in front of a partner and output you cannot comes down to a few habits, all of which the prompt enforces:
A market research brief that is actually useful tends to answer the same six questions, in roughly this order:
Definition. What is in this market and what is not. Sounds trivial, and it is where most disagreement about market size actually originates — two analysts sizing "AI infrastructure" differently are usually not disagreeing about numbers, they are drawing the boundary in different places.
Size and growth. TAM and the serviceable slice, the growth rate, and critically the provenance of each figure. Track the number back to who produced it and how.
Competitive map. Who is operating here, what each one actually does differently, and where the coverage is thin. Positioning claims from company websites are the least reliable input available; what a company hires for and what its customers say in reviews are far better signals.
The buyer. Who holds the budget, what they are trying to accomplish, and what they complain about. Review sites, support forums and community threads outperform any report for this, because they are the only place buyers describe the problem in their own words.
Recent change. What moved in the last twelve months — funding, new entrants, regulation, a shift in how buyers are procuring. Markets are usually described in a static way and behave dynamically.
The counter-case. What a skeptic would say. Structural headwinds, incumbents who could close the gap trivially, demand that looks real but is a temporary effect of something else.
Point a chatbot at a market and it will produce something resembling the above. The gap is in the sourcing discipline and the breadth of what gets read. Sai works across sources rather than summarizing the first page of results, cross-references figures that appear in multiple places to see whether they are independent or all tracing back to one original, and produces a document with the citations attached — so anyone reviewing it can check the claim that matters to them without redoing the research.
The output lands as a structured document with proper headers, which matters more than it sounds. Research that has to be reformatted before it can be used tends not to get used.