The team we hired in year one, honestly

A counterpart to the full-year retrospective on shipping grounded AI. That post is about the product thesis. This one is about the team we built around it. Shorter, less polished, more specific — not advice, just an account.

I am keeping numbers out of this post. Headcount specifics and comp specifics are not what I want to write about in public. What I want to write about is the shape of the hiring decisions, which roles we got right, and which ones I would sequence differently if I could.

What we hired

In rough order, the functions we staffed in year one:

• Engineering, specifically people with strong RAG and retrieval backgrounds rather than generalist AI hires.
• A proposal-craft lead — the real version of the voice the Sarah byline is a composite of. Someone who had lived inside a proposal function and could tell me where we were building abstractly versus where we were building what a proposal team would actually use.
• Product, specifically someone who could translate between engineering and the proposal-craft lead without either function feeling ventriloquized.
• Design, a single full-stack designer who owns both the product surface and the marketing site.
• Research, a part-time role that shipped the industry benchmarks and teardowns the Research byline published through the year.
• Growth, late in the year.
• One customer-success hire, very late in the year, once we had enough customers that the founder covering CS personally was producing a worse experience than a dedicated person would.

What I got right

Proposal-craft hire, early. Hiring someone who had shipped real proposal responses before we were deep into product design was the single best staffing decision of the year. Every product debate we had about how writers actually work, how SMEs actually interact, how color-team reviews actually run, had a voice in the room with evidence. Without that voice, we would have built for the proposal function we imagined, which is not the proposal function that exists.

RAG-specific engineers. AI engineering has become a broad enough field that “AI engineer” can mean many things. We hired specifically for people who had shipped retrieval-augmented systems, had opinions on chunking and reranking, and could tell the difference between a retrieval quality problem and a prompt engineering problem. That specificity paid off in the speed we could move on the retrieval eval pipeline and the chunking ablations — neither was debate-heavy because the team had priors.

Design as a full-stack role, not two hires. One designer owning both the product and the marketing surface meant the voice of the product matched the voice of the blog, which matched the voice of the /compare/ pages, which matched the voice of the pricing page. Most early-stage companies split design too early and end up with two different products with one name. We avoided that. I do not think we can keep it as a single-person role forever — we are already feeling the ceiling — but for year one it was right.

What I should have hired earlier

Customer success. For most of the year, I was the primary contact on our largest customers. That is fine when you have three customers and terrible by the time you have a dozen. The specific thing it cost us: I was doing CS work in the windows I should have been doing capture-side conversations with our prospect pipeline, which meant our sales cycle was longer than it should have been and our CS quality on existing customers was spikier than it should have been. The hire I made in Q4 should have happened in Q3 or even Q2.

The lesson, such as it is: the cost of founder-as-CS is not the hours the founder spends; it is the hours the founder does not spend on the next thing. I missed that tradeoff for longer than I should have.

A dedicated KB/content-ops person on our own team. We run our own proposal function internally — we use the product to respond to our own RFPs and DDQs, which has been a useful forcing function. What we did not have, for most of the year, was someone who owned the hygiene of our own KB. The KB rotted in the same way customer KBs rot — see the SME collaboration piece for the mechanism — and the signal got muddled by our own bad inputs. A part-time KB owner inside our own company would have been high-leverage. We now have one.

What I should have hired later, or not yet

Growth, in the shape we first tried it. We hired a growth role in H2 that was structured around outbound — lead generation, cadence, pipeline creation. It was a reasonable decision on paper and it did not work. The market we are in is small enough that the leads worth having surface through the blog, through the comparison pages, and through the research the Research byline publishes. Cold outbound produced meetings with prospects who were not ready to evaluate and did not convert.

We restructured the role mid-quarter toward what the market actually rewards: content distribution, analyst relations, and community work. That is working. The hire itself is still in seat and doing well in the restructured shape. What was wrong was my spec, not the person. I was hiring for a motion that did not fit a technical-credibility-driven category, and I had to learn that by spending six weeks watching a thoughtful person try to do the wrong job.

Senior IC hires we did not make. The temptation at a Series-A-shaped company is to hire senior ICs early — a principal engineer, a staff PM, a senior designer. We resisted most of those. The ones we did make, we made for specific scope reasons (one senior ML engineer, hired because the retrieval work needed their depth). The ones we did not make in favor of mid-level hires — I do not regret. Year-one PursuitAgent did not have the operational structure to make senior ICs productive, and a senior hire who does not get leverage in their first quarter leaves feeling underused.

The roles that are not on the list

Sales. We did not hire a salesperson in year one. I did the sales, the proposal-craft lead helped on product-side conversations with prospects, and the growth role in its restructured shape supports both. Whether that is sustainable into year two is an open question. The honest version is that I do not know yet.

Marketing. We did not hire a marketing lead either. The blog and the comparison pages have been the marketing function. That is structurally unusual and it works because the blog is a serious field journal and not a content mill — see the company voice doc for the stance. Whether to keep the marketing function owner-less into year two is, again, an open question.

Two things that are true about every year-one staffing decision

The first: you do not have the information you would need to make the right decision, and you have to make the decision anyway. The information comes in after the decision, and by then the next decision is already due. This is not a problem to be solved; it is the shape of the job.

The second: the specific people you hire matter more than the titles. We had two hires where I would have said in advance that the function was right and the person was wrong, and the person turned out to be right and the function turned out to be wrong. We had zero hires where the function was right, the person was wrong, and it got salvaged. The lesson is: when in doubt, hire the person and let the function adjust around them. The opposite path — hiring for the function and expecting a given person to grow into it — has a worse success rate in my experience.

What year two looks like

A CS hire earlier in the cycle, sized for the customer count we will have by mid-year. Probably a first sales hire, structured as a sales engineer with proposal-side credibility, not a quota-carrying AE in the traditional shape. A second engineer with deep ML infra experience to take on retrieval-tier work we have been deferring. A KB/content lead for the internal team. Probably not a marketing hire this year; we will see.

I was wrong about a few things this year. I expect to be wrong about a few more next year. The goal is to be wrong about different things each time.