AI FIELD MANUALEXECUTIVES, CONSULTANTS
How to build a defensible AI-competitive-landscape brief for a board meeting without an LLM quietly inventing a competitor's product, partnership or funding round.
Last reviewed 1 September 2026
A CFO at a mid-sized UAE logistics firm is told on Monday that Friday's board meeting needs a slide on "where competitors are with AI" - three named competitors, what they have shipped, and what the risk is of doing nothing for another year.
The obvious move is to ask ChatGPT or Claude "what is Competitor X doing with AI" and paste the answer into a slide. This is also the fastest way to put a fabricated partnership or an outdated executive quote in front of a board, because a general-purpose chat model answering from memory will produce a fluent paragraph whether or not the underlying facts are current or true.
The actual problem is not writing the brief. It is building one where every sentence traces to something the board (or a sharp director) could click through and verify.
Separate research from writing. Do the fact-gathering in a tool that searches the live web and cites its sources as it goes - not a plain chat model working from training data - then only hand the LLM the verified facts to structure into board language.
Treat the LLM as an editor of evidence you already collected, never as the source of the evidence itself. Anything the model adds that isn't in the source material gets deleted, not fact-checked after the fact.
A board brief fails for one of two reasons: it says nothing decision-useful, or it says something wrong. The workflow above targets the second failure specifically, because it is the one that ends careers - a slide with a stale but harmless observation is forgettable, a slide with an invented funding round is not.
Search-grounded tools attach a citation to each claim at generation time, which turns fact-checking from "research the topic yourself" into "open ten links and confirm they say what the slide says" - a task that fits in the hour you actually have before Friday.
Before touching a tool, write down what the board actually needs to decide: usually some version of "are we behind", "what would catching up cost", and "what is the downside of waiting another year". A brief built to answer three questions is checkable; a brief built to "cover AI" is not.
Use a search-grounded assistant (Perplexity, or Claude/ChatGPT with browsing switched on) with one prompt per competitor: what AI capability have they publicly announced or shipped in the last 12 months, with a link for each claim. Keep the raw citations, not just the summary.
For each claim the tool returned, open the source. Press releases and the vendor's own product pages count; a blog aggregator restating a press release does not add anything and often drops the original date. Discard any claim you cannot trace to a primary source.
Before any slide copy, build a plain table: competitor, capability, source, date. This is the artifact that survives a hostile question in the room - if a director asks "where did that come from", you have the row.
Paste the finished table into Claude or ChatGPT and ask it to draft board-register slide copy from exactly those rows - explicitly instructed to add no fact not present in the table. Cut anything that reads like it needed a citation but doesn't have one.
The table alone is not a recommendation. Close with one paragraph of your own judgment - written by you, not the model - on what the pattern across the three rows implies for this specific company.
Search-grounded research with inline citations for the competitor fact-finding pass.
www.perplexity.ai (opens in a new tab)Drafting board-register slide copy from an already-verified fact table, and browsing for the initial search pass.
claude.ai (opens in a new tab)Alternative for both the browsing research pass and the drafting pass.
chatgpt.com (opens in a new tab)Search-grounded tools still occasionally cite a source that does not actually support the claim next to it - the citation existing is not proof the claim is correct. Opening every link is not optional.
Public information has a lag. A competitor's real AI posture (internal pilots, procurement in flight) is often a quarter or two ahead of anything publicly announced, so a brief built this way describes what is announced, not what is true - say that limitation out loud in the room.
This method is weak on private companies with little public footprint. If a named competitor has no real coverage, the honest slide says so rather than padding the gap with an inference dressed as a fact.
Advertisement
A Dubai-based freight forwarder needs a board slide on how three named regional competitors are using AI in customer-facing operations.
Perplexity query per competitor: "[Competitor] AI customer service OR AI logistics tracking announcement 2025 2026", citations kept.
Of nine total claims returned across three competitors, two were unsupported by their cited link on inspection and were dropped.
The surviving seven claims went into a fact table: competitor, capability, source, date.
Claude drafted three slides from the table, one per competitor, instructed to add no unsourced claim.
The CFO added a closing paragraph: two of three competitors have shipped a customer-facing AI feature in the last nine months; the firm has shipped none; the cost of a six-month pilot is a smaller number than the cost of a second year of no visible movement.
A five-slide brief where every competitor claim has a live link behind it, presented without a single question the CFO couldn't answer with "here's the source."
Nakoda editorial · last reviewed
This entry describes a workflow Nakoda recommends - it is not a claim about how any named tool behaves in every case, and it is not paid placement. Spotted something out of date? Tell us.
Workflow
A source-first market-sizing method - AI structures the framework, but every figure on the slide traces to a named, checkable source, never a model's confident-sounding estimate.
