An AI Search Employee is an AI agent given the goal of getting a company recommended in AI answers, the way a hire would be: it learns who the company sells to, finds what AI engines are missing about it, writes the work for approval and reports what moved. Obi, by Heysurface, Inc., uses the name for its role.
The name describes a way of working more than a kind of software. An AI Search Employee takes the outcome (being named in the answers a company's buyers read) and does the work toward it: it studies the business, decides what to do first, produces the pages and fixes, and reports on the result. AI visibility dashboards stop earlier, at measuring where a brand appears, and leave the work to the team.
Obi, the AI Search Employee made by Heysurface, Inc., runs as a CLI and a Skill inside the coding agent its customer already uses (Claude Code, Codex, Cursor and other agents, or any terminal). It proposes the buyer segments of the ICP, asks AI engines the questions those buyers ask, finds the missing pieces, writes strategies and pages in the company's voice, and hands every page over as a draft. AI Search Agent is the other name used for the same role.
How it works
The work runs in rounds. The employee reads the company's website and what the founder tells it, proposes up to four buyer segments for confirmation, writes the questions those buyers ask, asks the AI engines those questions, then proposes the next move by how close each question sits to a purchase. Each round shows what moved since the last.
Why it matters
Most AI visibility tools stop at measurement, and the work that changes an AI answer (a new comparison page, a fix that lets crawlers read a page, an outdated third-party fact corrected) stays on the team's list. An employee model is built to close that gap: it starts from the buyers who sign and produces the work, so a small team can act on AI search without hiring a specialist.
Example
A five-person B2B startup installs Obi in Claude Code, confirms two buyer segments, and receives a draft page answering the comparison question its buyers ask ChatGPT, ready to review and publish.
ICP-Based AI Search is an approach to AI search that starts from the ideal customer profile (ICP): the buyers who sign, their situation and their words. The questions measured come from that profile and are ranked by how close each sits to a purchase, so the work aims at recommendations that can become revenue.
Much AI visibility tracking starts from keywords, or from prompts anyone might type. ICP-Based AI Search starts from the ideal customer profile: who buys, in what situation, and how they phrase their questions. In Obi's version, the agent reads the company's website and what the founder tells it, then proposes up to four segments, each a buyer and the situation they are in. The founder confirms the segments before any question is written.
From there, every question is written the way a buyer in that segment would ask an AI, and each is ranked by its likely impact on a purchase. Visibility on a question nobody in the ICP asks counts for little; a recommendation on the question a qualified buyer asks before a demo counts most.
How it works
Segments first, confirmed by the founder; then questions per segment, in the buyers' words; then a ranking by buyer impact that orders what is measured and what is worked on.
Why it matters
Being mentioned on generic prompts brings generic traffic. A startup with a narrow market needs the AI answers its own buyers read to name it, which means measuring and working on those questions first. Starting from the ICP keeps the effort on the questions closest to a sale.
Example
A medtech startup confirms one segment, clinical operations leads at mid-size hospitals; the questions written are the ones those leads ask ChatGPT about patient intake software, ranked by how close each is to a buying decision.
Also called: contextual buyer prompt, buyer prompts, contextual prompts
DocumentedOrigin: Obi's method (Heysurface, Inc.)
Contextual Buyer Prompts are the questions asked of AI engines to measure a company's visibility, written the way a buyer in its ICP would ask them, with that buyer's words and situation. They replace prompts built from keywords, so what is measured matches what real buyers read.
A prompt built from a keyword ("best CRM software") measures a question few real buyers type into an assistant. A contextual buyer prompt carries the buyer's situation and vocabulary ("Which CRM works for a five-person agency that already uses Notion and needs client portals?"), because assistants answer long, specific questions differently from short ones.
Obi writes these prompts for each confirmed segment of the ICP, from the words those buyers use, and ranks each by buyer impact. They are the questions it asks the engines every round.
Why it matters
AI answers change with the context in a question, and assistants expand a question into many background searches before answering. Measuring short keyword prompts can show a brand as visible on questions its buyers never ask, and invisible on the ones they do. Writing prompts from the buyer's context makes the measurement and the work point at the same answers buyers see.
Example
Instead of tracking "patient intake software", the prompt tracked is "What patient intake tool integrates with Epic for a 200-bed hospital with a small IT team?"
A missing piece is first-hand evidence or an explanation that a company's buyers need and that AI cannot find anywhere yet, such as a page on how a product fits a five-person team. AI engines can only cite what exists, so creating missing pieces from a company's own data adds what no other source says.
When an assistant answers a buyer's question, it can only cite and recommend what it finds. If the facts that would make a company the right answer exist only in sales calls, support tickets or the founder's head, the assistant answers without them. A missing piece is one of those facts or explanations, worth creating because nothing public covers it.
Missing pieces are first-hand: a measured result, a specific fit, a comparison the company can make honestly, an explanation only its team can give. Neither uniqueness nor an access fix guarantees a citation, a recommendation or revenue; a missing piece raises the odds by giving the engines something they need and cannot find elsewhere.
Why it matters
Rewriting what already ranks adds little an engine has not seen. The pieces that change an answer are the ones that add information: the gap between what buyers ask and what the web says about a company. Finding that gap from the buyers' questions, and filling it with the company's own data, is the work most likely to earn a recommendation.
Example
Buyers ask whether a scheduling tool works for clinics with rotating shifts; no page anywhere answers it, so the company publishes its own rotating-shift setup guide with numbers from three customers.
Missing-Piece Content is content written to supply a missing piece: a page built from a company's own data, in its voice, that answers what buyers ask and AI cannot find elsewhere. It is shaped to be useful to visitors and easy for AI engines to read and quote, and each page is approved before it goes live.
Missing-Piece Content turns a missing piece into a page. It is drafted from the company's proprietary data and in its voice and tone, with a direct answer early, the evidence after it, clear headings and dates, so a visitor can use it and an AI engine can read and quote it.
In Obi's method, the company reviews and publishes; Obi publishes nothing on its own. Its CLI offer covers strategies and website work; publishing on Reddit, review sites or the press is outside it.
Why it matters
Generic content repeats the web and gives an engine no reason to cite it. A page that carries first-hand data, written to answer a real buyer question, gives the engine a source it lacks. Keeping the company's voice and its approval in the loop keeps the page accurate and on brand.
Example
From a support export, a draft page on migrating from a competitor in under a day, with the five steps customers actually took; the founder edits one paragraph and publishes it.