01
Segment
Now
Product startups with a mature product (at least a year of development) and at least 4–5 engineers.
Actual
Product companies with 50–100 engineers. AI-forward, with heavy adoption and considerable usage.
Agencies, outsourcing, and teams small enough that the whole product fits in one agent context are out.
02
Buyer and user
Now
Top down
Buyer: CTO (usually a user as well)
User: Programmers
Actual
Bottom up
Buyer: Engineering manager (can approve the purchase without a formal procurement process, land with his team, expand organization-wide), head of engineering, head of AI, head of technology, and similar roles
User: Programmers, EMs, PMs
03
Problem
Coding agents already have access to the apps where a company describes how it works, how the product works, and how it is implemented, including Slack, GitHub, and Jira. Finding the correct information there is still difficult.
Slack has no semantic search. If the agent does not know the exact words, and the engineer does not point to the thread with a link, it will not find those messages. The same is true for tickets and for code. The agent might find them, but only by exploring the whole repository, or the whole GitHub organization.
The models were trained on the public internet, and the specific ways a company works are not in that data. A lot of what happens, including meetings, is never written down anywhere. Documentation goes out of date, and it goes out of date faster as the pace of AI development increases.
These do two things. Exploring that much costs tokens. If the agent does not explore, it goes with assumptions, which makes a mistake more likely, which costs again, or worse, it passes the check and ends up wrong in production.
04
Value proposition
Nucipal makes coding agents perform better by removing the assumptions they make, which reduces errors, which reduces cost. (We do not have the usage data required to put this into numbers.)
05
Positioning
Nucipal is a provider-agnostic infrastructure layer that any agent can call. It is automatic, so there is no manual input of data and no manual upkeep. It is tailored to small and medium businesses, not enterprise-level contracts.
| Nucipal | Is the alternative to |
|---|---|
| Infrastructure other agents call, updated automatically from the apps used for everyday work | Knowledge typed into Notion by hand, dug out of Slack, or kept only inside ChatGPT |
| The same job for a smaller company, at a price that company can approve | Glean, and other enterprise search, where contracts sit around 60,000 euro |
06
Channel
Now
LinkedIn founder outreach to fellow startup founders and CTOs we know, or through a mutual connection.
Actual
Cold email to engineering leadership (head of engineering, engineering manager, head of AI, head of technology, and similar roles).
07
Conversion
The first call is thirty minutes of discovery. We see if they fit our segment and if the problem shows up in their work. The second call is a demo, and that is when we ask for a four-week design partnership pilot.
We don't have defined success criteria yet. One example is: more than half the time the agent calls Nucipal, the answer is useful.
After four weeks, if the success criteria is met, we discuss a contract. If not, we discuss extending the duration of the pilot to improve the product.
Now
LinkedIn → discovery call → demo → four-week pilot
Actual
Cold email → discovery call → demo → four-week pilot on one engineering manager’s team → contract and org rollout
08
Evidence loop
Customer profile
Jobs, pains, and gains
Now
Get an agent to implement the way this company builds.
Speed and maintenance overhead. Shipping is very fast, so stopping to document how systems work means falling behind. The information lives in engineers’ heads or in private agent sessions, so it is forgotten or never read again.
The team keeps moving, does not have to document anything by hand, spends less time re-prompting and redoing the work, and the agent still has the knowledge.
Actual
Get an agent to implement the way this company builds.
Fragmentation. Teams in a large organization own different parts of the system. Working across those parts depends on someone stopping their own work to help. Agents cannot ask for that help.
An engineer can implement against another team’s rules without pulling that team in.
Business model
One customer, one offer, one channel.
| Customer | Product startups with 4–5 or more engineers, then companies with 50–100. |
| Offer | Better agent performance and lower agent cost, with no wiki to maintain. |
| Channel | LinkedIn for the first startups. Cold email for companies with 50–100 engineers. |
| Relationship | The founders onboard the team and sit in a weekly session through the pilot. |
| Revenue | Design partnership is free. Payment starts after the pilot shows a gain in performance or cost. |
| Cost | Model usage and infrastructure for ingesting company context and serving it to agents. |
Design partners
Two or three from the Now segment.
Product companies we can reach directly: a mature product (at least a year of development), at least 4–5 engineers, AI already in the work, and the same gap between their agents and how the company actually builds.
From each partnership we want to learn how agents get that company’s knowledge today, what they already tried, what it costs, and who besides the first contact would use a fix.