# Your AI starts every chat from zero. Here's the fix for your team (https://www.hyreflow.ai/blog/your-ai-starts-every-chat-from-zero) By Niklas Huetzen, Co-founder, hyreflow. Published 27 Oct 2026. > Claude and ChatGPT forget your business between chats, so every recruiter on your team explains the same clients, roles and tone again. hyreflow's second brain fixes that with shared notes on your company, candidate profiles and client profiles, built from your website. Every teammate's AI assistant reads them before it starts. When you tell it what worked, it offers to save it, and your team decides what it keeps. - AI assistants carry no memory between conversations. Each new chat meets your business for the first time. - A personal memory stays with one person. A recruitment team needs its knowledge where every consultant's assistant can reach it. - hyreflow builds the first version of your second brain from your website, LinkedIn page and jobs board, and every note says where it came from. - Your assistant suggests what to remember. Nothing changes your team's notes until a person confirms it. - The more your team confirms, the better every consultant's assistant knows your business. Your AI assistant is brilliant and has never heard of your agency. Every new chat in Claude or ChatGPT starts with a blank page: no clients, no desks, no idea who you place. hyreflow's second brain gives that assistant a memory of your business, shared by your whole team, so no chat starts from zero again. ## Why does your AI start every chat from zero? [#why-does-your-ai-start-every-chat-from-zero] AI assistants start every chat from zero because the models behind them keep no memory between conversations. Anthropic's documentation says plainly that its Messages API "is stateless", so the full conversation has to be sent every time.[^1] OpenAI describes each request the same way: "independent and stateless".[^2] For a recruiter, that means briefing your assistant every morning. You paste the client notes again. You explain again that you place platform engineers, not SREs. You describe the tone your outreach should have, for the fifth time this week. Now multiply that by a team. Three consultants describe the same client three different ways. One prompt finds senior hires, another finds juniors, and the outreach reads like three different firms. The assistant is not the problem. It just never got the brief. Without a second brain With a second brain Monday chat Who do you place? What is the role? Tuesday chat Who do you place? How should it sound? New consultant's chat Who do you place? Which clients? Second brain Monday chat starts from your overview Tuesday chat starts from your overview New consultant starts from your overview The assistant is not the problem. It just never got the brief. ## What is a second brain for a recruitment team? [#what-is-a-second-brain-for-a-recruitment-team] A second brain for a recruitment team is a shared set of notes about your business that every consultant's AI assistant reads before it does anything. The idea comes from Tiago Forte, who described a second brain as "an external, centralized, digital repository" for what you learn.[^3] hyreflow applies it to the assistant rather than the person: the notes are written for your AI to read and use. The notes hyreflow keeps about your business: your company, the candidates you place, the clients you serve, and what wins them. Your AI assistant reads them before it asks you anything, and everyone in your workspace shares the same notes. Why a team needs this more than a person does: personal memory stays personal. In Claude, for example, memory is set per person, and workspace owners "can't view or edit a user's individual memories".[^4] That is the right design for a private assistant. It also means what one consultant teaches their assistant never reaches the desk next to them. A recruitment firm's knowledge lives in the team, not in one chat history. hyreflow's second brain puts it where every teammate's assistant can reach it. ## How does hyreflow build your second brain on day one? [#how-does-hyreflow-build-your-second-brain-on-day-one] hyreflow builds the first version of your second brain from your company's public footprint while you finish signing up. It reads your website, your LinkedIn company page and your jobs board, and writes the first notes in a few minutes. Two rules keep the first build honest. Every note says where it came from, and a note hyreflow cannot back with a source is left out. When hyreflow is unsure, it marks the note as a draft, and your assistant checks it with you before relying on it. From then on, every chat starts with an overview of your business instead of a blank page. The [second brain docs](/docs/knowledge) cover the details of the build. ## What does hyreflow's second brain know about your business? [#what-does-hyreflows-second-brain-know-about-your-business] hyreflow's second brain holds four kinds of knowledge: who you are, who you place, who you sell to, and what wins. Each note is short, and notes link to each other, so a candidate profile leads straight to the pains, angles and proof that belong to it. * **Your company.** What you do, what you offer, and your desks. * **Candidate profiles.** The people you place, each with the search filters that find them: titles, locations, skills, keywords and seniority. * **Client profiles.** The companies you sell to and the people who buy from you. * **What wins.** The pains each profile feels, the triggers that make them move, the outreach angles that work, and the proof points you can use. hyreflow sorts these notes into folders. An agency gets every folder. An in-house team gets the same brain without the desks, offers and client folders, because it hires for one company: its own. | Folder | What a note in it holds | Links to | | ------------------ | ------------------------------------------------------------------------------- | --------------------------------------------- | | Company | What you do, where you are, how big you are | | | Desks | Each market or specialism you recruit for (agencies) | | | Offers | What you sell clients, such as contingent or retained search (agencies) | Its desk | | Client ICP | The kind of company that buys from you (agencies) | | | Client personas | The people at those companies who hire you (agencies) | Their client ICP | | Candidate personas | The people you place, in plain words | Their desk | | Search filters | Titles, locations, skills, keywords and seniority that find a candidate persona | Its candidate persona | | Pains | What frustrates that candidate in their current job | Its candidate persona | | Triggers | What makes them open to a move | Its candidate persona | | Angles | An outreach angle that speaks to them | Its candidate persona and the pain it answers | | Proof | Placements and results you can point to | Its candidate persona | ## How do the notes in a second brain connect? [#how-do-the-notes-in-a-second-brain-connect] The notes in hyreflow's second brain connect like a small map of your business. Every pain, trigger, angle and proof point hangs off the candidate persona it belongs to. At an agency, each candidate persona sits on a desk. Every candidate persona carries its own search filter, so a search can start straight from it. Offers Contingent search Desks Platform engineering Client ICP Series B fintech Search filters Titles, skills, seniority Candidate personas Senior platform engineer Client personas VP Engineering Pains On-call burnout Triggers New funding round Angles Calmer on-call shifts Proof Scale-up placements Dashed box: agencies only. Dashed line: the angle answers this pain. The links are what make the second brain useful to an assistant. When you ask for outreach to senior platform engineers, your assistant opens that persona, follows its links to the pains, angles and proof, and writes from all of them. When you ask for a search, it follows the link to the search filter instead. Every note also lists the notes that link to it, under **Linked from**. hyreflow works that list out for you, so a new angle shows up on its persona's page the moment it is saved. A link to a note that does not exist yet starts working as soon as someone adds that note. ## What does one note in the second brain look like? [#what-does-one-note-in-the-second-brain-look-like] One note in hyreflow's second brain is a short page in plain words, with its folder, its status, its links and its sources. It is written to be read in seconds, by you or by your assistant. Candidate personas Active Senior platform engineer Engineers who run cloud infrastructure at scale-ups, usually several years in. They move for calmer on-call and better tooling. Platform engineering Search filter Sources your-company.com/careers linkedin.com/company/your-company Linked from On-call burnout Calmer on-call shifts Scale-up placements Built from your site, edited since Folder and status draft when hyreflow was unsure A short note in plain words Links to related notes Where each fact came from Notes that link here worked out for you How it was made, and when Each note carries one of three statuses: * **Draft.** hyreflow was not sure, so your assistant checks the note with you before relying on it. * **Active.** Confirmed or well sourced. Your assistant uses it without asking. * **Archived.** Taken out of use. It no longer shows up when your assistant searches, and stays on the Knowledge page for reference. The **Knowledge page** in your hyreflow dashboard is where you see all of it. It lists every note, with a search box and filters by folder and status, and a map view of how the notes link. Open a note to read it with its sources and its Linked from list. Each note also says whether it was built from your site, added by an assistant, or edited since, and when. ## How does your AI assistant use the second brain? [#how-does-your-ai-assistant-use-the-second-brain] Your AI assistant reads the second brain before it asks you anything about your business. If you keep your own ICP document, that comes first. The second brain comes next, and your assistant only asks you when neither has the answer. The difference shows on the first chat. Instead of asking what you recruit for, your assistant says what it found: "hyreflow read your-company.com, and the person you hire for most looks like a senior platform engineer in London. Right?" You confirm, and it goes straight to a search built from that profile's filters. Your assistant finds what it needs in three moves. It first reads the list of notes, like a table of contents. It then searches for the topic at hand, such as a client or a role. Finally it opens only the notes it needs and follows their links. It works the same way for everything after. A shortlist starts from the right candidate profile. A business development list starts from your client profiles. Outreach starts from the angles and proof your team already agreed on. Prompt (Claude): > Show me what hyreflow learned about my business. What comes back: A summary of your second brain: your company, the candidate profiles it found with their search filters, your client profiles, and any notes still marked as drafts for you to check. ## How does the second brain learn without guessing? [#how-does-the-second-brain-learn-without-guessing] hyreflow's second brain learns from what your team tells it, not from guesses. When you tell your assistant what worked or what changed, it offers to file it: it shows you the note it would write, and saves it once you say yes. Second brain shared notes Read before it asks Work search, BD, outreach Suggest you said what worked Confirm a person says yes Say you tell your assistant: "We don't place SREs, we place platform engineers." It shows you the candidate profile with the change and asks whether to save it. One sentence, and every consultant's next search uses the corrected profile. The same works for what you learn on the desk. An angle that got replies, a new client, a must-have a client keeps repeating: tell your assistant, and it offers to file it where it belongs. Why the confirm step matters for a team: nothing is saved until the person asking says yes, and the change shows on the Knowledge page. A note your assistant built as a draft becomes active the moment someone confirms it. To take a note out, ask your assistant to archive it. Archived notes stop showing up in searches but stay on the Knowledge page. ## How does one second brain keep a whole recruiting team consistent? [#how-does-one-second-brain-keep-a-whole-recruiting-team-consistent] One second brain keeps a recruiting team consistent because every consultant's AI assistant reads the same notes. The same candidate profiles, the same angles and the same proof points show up in every chat, so results stop depending on who wrote the prompt. * **One fix reaches everyone.** A correction one consultant confirms is in every teammate's next chat. * **New starters begin informed.** A new consultant's assistant knows your desks, clients and candidate profiles on their first day. * **Any assistant works.** Claude, ChatGPT and other assistants that [connect over MCP](/docs/mcp-setup) all read the same notes. * **Your notes stay yours.** Everyone in your workspace shares them, and no other workspace can read them. In-house talent acquisition teams get the same benefit. Every recruiter's assistant hires against the same profiles, so a hiring manager hears one consistent story about the role, whoever runs the search. ## What changes day to day when your AI never starts from zero? [#what-changes-day-to-day-when-your-ai-never-starts-from-zero] When your AI assistant never starts from zero, you skip the briefing and go straight to the work. You stop pasting client notes, describing roles and explaining your tone before every task. | Before the second brain | With the second brain | | ------------------------------------ | ---------------------------------------------- | | Brief the assistant on your firm | Your assistant already has the overview | | Describe the role and the filters | It starts from the matching candidate profile | | Paste the client's must-haves | They are in the client profile | | Explain how your outreach sounds | It writes from your confirmed angles and proof | | Every consultant prompts differently | Every assistant works from the same notes | A Monday [shortlist from a job description](/blog/how-to-turn-a-job-description-into-a-shortlist-with-claude) starts from the right profile. A business development list for a new desk starts from your client profiles, and a [search of your team's LinkedIn network](/blog/how-recruiters-should-use-their-linkedin-network) knows who to look for. Outreach sounds like your firm, not like whoever happened to write the prompt. Sign up with your work email, connect hyreflow to Claude or ChatGPT, and ask your assistant what it learned. ## Is this the step toward an AI assistant that grows with your business? [#is-this-the-step-toward-an-ai-assistant-that-grows-with-your-business] A shared second brain is the step from an AI assistant you brief every day to one that grows with your business. Today's assistants are smart strangers: capable, fast, and new to your firm every morning. The missing piece is memory, and for a team, that memory has to be shared. Every note your team confirms makes every consultant's assistant a little sharper. Corrections, client preferences and the angles that worked add up week by week. Your assistant in three months knows your firm far better than it did on day one. Day one A few months in Built from your site Added when your team confirmed That knowledge also stays with your firm. It no longer lives only in one consultant's head or one person's chat history. When someone leaves, what they taught the second brain stays for everyone who comes next. ## Five things to teach your second brain this week [#five-things-to-teach-your-second-brain-this-week] The fastest way to grow your second brain is to tell your assistant what you already know. These five prompts cover what matters most for a recruitment desk. "Check my candidate profiles in hyreflow. We place platform engineers, not SREs, mostly in London and Manchester." Your assistant shows the change and saves it once you confirm. "Add to the client profile for fintech scale-ups: they want people from regulated environments and never hire from big consultancies." "The angle about on-call burnout got three replies this week from platform engineers. File it as an angle for that profile." "Add a proof point: we filled four platform roles for one Series B client this year. We can mention it without naming the client." "Our outreach is short, direct and never uses exclamation marks. Save that as our tone." ## Where hyreflow's second brain falls short [#where-hyreflows-second-brain-falls-short] The first build reads only public sources, so it is a first draft of your business, not the full picture. A thin website gives a thin first version, and the notes fill out as your team adds to them. Changes happen through your AI assistant, not by editing on the Knowledge page. There is no undo: archive a note to take it out. The built second brain comes to new workspaces first; existing workspaces can already add notes through their assistant. ## Questions recruiters ask about the second brain [#questions-recruiters-ask-about-the-second-brain] ### Can other companies see my second brain? No. Everyone in your hyreflow workspace shares the same notes, and no other workspace can read them. ### Which AI assistants can use hyreflow's second brain? Any assistant connected to hyreflow over MCP, including Claude and ChatGPT. They all read the same notes, so a mixed team stays consistent. ### Does the second brain change anything without asking me? No. Your assistant suggests a note and shows you the change. It saves only what you or a teammate stated or confirmed. ### What if hyreflow got something wrong about my business? Tell your assistant what is wrong. It shows you the corrected note and saves it once you confirm. Notes hyreflow was unsure about are marked as drafts, so your assistant checks those with you first. ### I already keep my own ICP document. Which one wins? Your own ICP document comes first. Your assistant uses the second brain where your document has no answer, and asks you only when neither does. ### Does reading or writing notes cost credits? No. Your assistant reads, searches and updates your second brain without spending credits. Credits are spent on data lookups, such as searches and enrichment; see [credits and billing](/docs/credits-and-billing). [^1]: Anthropic, Using the Messages API, platform.claude.com. [^2]: OpenAI, Conversation state, developers.openai.com. [^3]: Tiago Forte, Building a Second Brain: The Definitive Introductory Guide, fortelabs.com. [^4]: Anthropic Help Center, Use Claude's chat search and memory to build on previous context. ## Sources - [Using the Messages API](https://platform.claude.com/docs/en/build-with-claude/working-with-messages), Anthropic. Checked 27 Oct 2026 - [Conversation state](https://developers.openai.com/api/docs/guides/conversation-state), OpenAI. Checked 27 Oct 2026 - [Use Claude's chat search and memory to build on previous context](https://support.claude.com/en/articles/11817273-use-claude-s-chat-search-and-memory-to-build-on-previous-context), Anthropic Help Center. Checked 27 Oct 2026 - [Building a Second Brain: The Definitive Introductory Guide](https://fortelabs.com/blog/basboverview/), Forte Labs. Checked 27 Oct 2026 - [Second brain](https://www.hyreflow.ai/docs/knowledge), hyreflow docs. Checked 27 Oct 2026