The problem this solves
Competitive briefs go stale the week after they are written, and the reason is always the same: they were assembled by one person reading sites on one afternoon. Positioning changes quietly. A pricing page gets a new tier, a comparison page adds a rival, a claim that used to be about speed becomes a claim about control. By the time a brief reaches a seller it describes a company that has moved on.
The other failure is that briefs tend to record conclusions instead of sources. Somebody writes that a competitor targets mid-market and prices per seat. Three weeks later a rep repeats it on a call, the prospect corrects them, and the credibility cost lands on the rep rather than the brief.
What a seller actually needs is narrower than a full market study: who does this company say it competes with, what does it claim against them, how does it charge, and what proof does it publish. Those four questions have public answers on a small number of pages, and they are worth re-reading rather than remembering.
How the mission runs
- Read how the company positions itself. The Company Intelligence Agent works through the site: the home page, the comparison pages, the pricing page and any customer proof. Reading the comparison pages first is deliberate, since a company naming a rival is the most direct statement of positioning available.
- Collect the rivals it names. Every competitor the company mentions on its own pages is collected, and the claim made against each is kept with the wording. This is the part a seller can use directly, because it shows the argument the company has already decided to have.
- Research the three closest competitors the same way. The same questions are asked of the nearest rivals, whether they were named or surfaced as lookalikes. Asking identical questions of each is what makes the answers comparable instead of four essays in different shapes.
- Cite every claim. Each answer carries the page it came from and the wording that supports it. A brief where every line can be opened is a brief a seller will quote from, and one where a prospect disagreeing simply means checking the link.
- Deliver a comparison brief. Findings are assembled into a document in Google Docs, with a chart where the pricing models are comparable enough to show side by side, and a short summary of where the positions genuinely differ.
The prompt
This is the exact objective the agent receives. Swap the obvious placeholders for your own domain, segment or channel and run it as-is from the console, Slack, or the API.
What comes back
A comparison brief in Google Docs covering the company and its three closest rivals: named competitors, the claims made against each, pricing model and published proof, with a source link on every line and a chart where the models line up.
Make it yours
- Run it across a list of accounts rather than one company when a whole segment needs the same treatment.
- Narrow the questions to pricing alone when a deal is stuck on commercials.
- Schedule it quarterly against the same companies and keep each version, so a change in positioning is visible as a change.
- Add a question about which integrations each company leads with when the market competes on ecosystem rather than features.
Frequently asked questions
Is this different from a lookalike search?
Yes. A lookalike search finds companies that resemble one another on firmographics. This reads how a company argues for itself, which is a question about published language rather than company shape.
What if a company names no competitors?
That is a finding in itself, and the brief says so. The three closest rivals are then drawn from lookalikes so the comparison still has a shape.
How current is the brief?
It reflects the pages as they read when the mission ran, and every line links to the page. Running it on a schedule and keeping the versions is how teams turn that into a view of how positioning moves.
Can it cover private companies with thin sites?
It covers what is published. Where a company says little, the brief is short and says which questions had no public answer, which is more useful than filling the gaps with inference.