A consulting firm built for
the year consulting changed.
NexcurAI was founded on a simple observation. Most consulting work is either done badly by people who know too little, or done slowly by people who know plenty but cannot clone themselves. Claude changes that math. We are building the firm that takes the second group's judgment, pairs it with the model's scale, and ships artifacts that actually survive in the client's hands.
The founder & operator.
I spent the first ten years of my career writing code and the last five breaking it. The pattern I kept seeing was the same in both halves: companies know what good looks like in the abstract, they just cannot afford the hours it takes a senior operator to produce it.
When Claude's capabilities crossed the threshold where a careful person plus a careful model could produce senior-grade work at a fraction of the cost, I kept waiting for the firm that would operationalize it. No one did. The AI agencies spun up around unattended pipelines and cheap volume. The legacy firms treated AI as a demo slide. The gap was obvious.
So NexcurAI is what I wish I could have hired during my own engagements on both sides of the table. Senior operators. Claude used in the plumbing, not as the product. A long-form artifact at the end. No theater, no billable hours games, no we will get back to you next quarter.
We are small today, deliberately. The next hires are operators, not salespeople. The roadmap is the client list.
Nine things we hold to.
These are the beliefs that shape which clients we take, how we use the model, what we ship, and what we refuse to do even for a paying client. They are written down so you can hold us accountable.
Humans own the judgment.
Claude drafts, correlates, and critiques. A senior operator decides what ships. The model never signs off on a deliverable.
Ship the document, not the deck.
Every engagement ends with a long-form handbook a new hire can use. Slide decks are summaries of the handbook, not the other way around.
Price honestly, scope precisely.
Fixed-scope engagements. No hourly meters. If we miscalculate on a fixed engagement, the overrun is our problem, not yours.
Say no to bad fits.
We have turned down engagements worth six figures because they were not a fit. We will keep doing it. Scope creep starts at the pitch.
The model makes mistakes. Log them.
Every time Claude hallucinates or misses a detail, we log it in our eval harness. Our prompts improve monthly. Our mistakes get less common over time, not more clever.
Security is a first-class citizen.
The founder is a security person. We run our own infrastructure with the same rigor we demand from clients. We publish our security posture.
Transparent about Claude.
Every engagement tells clients what parts were drafted by Claude and what parts were authored by humans. We do not hide the model, and we do not hide behind it.
Refuse weapons, espionage, dark patterns.
We will not help offensive capabilities against unconsenting parties. No dark pattern growth. No deceptive marketing. Written into our SOWs.
Small team, high trust.
We grow slowly. Every operator is someone I would trust with my own production systems and my own reputation. That is the hiring bar.
Operators with leverage.
A look at the actual operating model. Not a vision. What we do every week.
The team
Today: one founder-operator (security and product lines), one senior engineer consulting part-time, one editor contracting on handbooks, one founder-advisor on marketing. By end of 2026, we plan to be four to six full-time operators across the five service lines.
We will not grow faster than the operator bar allows. If that means we turn away clients for a quarter, we turn away clients for a quarter.
The tooling stack
- ModelClaude Opus & Sonnet via API, with prompt caching, extended thinking, tool use
- TemplatesNexmill, our versioned handbook template library
- Eval harnessCustom, runs nightly on all prompts
- DeliveryGitHub repos per engagement, Notion for client-facing, PDF for archival
- CommsShared Slack channel or Teams, Linear, 30-min weekly live
- Security1Password, YubiKey for all operators, isolated client secrets vault
The first twelve months.
We are writing this in the open so future clients, partners, and hires can see the shape of the plan and call us out if we drift.
Founding quarter.
Incorporate. Build the first template library (security and web lines). Ship three inaugural engagements at discount to build case studies. Publish the first ten pieces of writing. Set up eval harness and internal tooling. Target: $150k revenue, three handbooks shipped.
Standardize and scale the framework.
Complete template library across all five service lines. Bring on first full-time operator (product). Launch the Share-of-Answer SaaS. Publish handbook sample publicly. Hit 8 active engagements. Target: $400k cumulative revenue, five handbooks shipped.
First retainer cohort.
Land 6 monthly retainers (mix of security, SEO/GEO, marketing). Bring on second operator (security). Publish Nexmill templates as a licensed product. Build out public case studies. Target: $950k cumulative revenue, 10 handbooks in circulation.
$2M ARR run rate.
Hit $2M annualized run rate. Team of four full-time plus two part-time. First enterprise retainer. Operate profitably without outside funding. Begin the second year plan with a clear data-backed view of which service lines to double down on.
Things we will tell you even if we don't have to.
Our operating numbers
We publish an anonymized operating update quarterly: revenue, engagement count, utilization rate, client NPS, and the failure modes we hit. If you are considering becoming a long-term retainer client, you deserve to know whether we will still be here in 18 months.
Our conflicts
We disclose every conflict at the scoping stage. If you are a competitor of an existing client, we tell you. We keep a public-facing client list (anonymized by industry if they wish). We do not work for both sides of a competitive matchup.
Our mistakes
When we get something wrong, we write it up. Internal retrospectives on missed findings and bad calls are shared with the affected client, and the patterns are worked into our templates and evals. The learning compounds, but only if we write it down.
Our limits
We are not the right fit for pure creative work without a strategic frame, for commodity development without an operator layer, or for clients who want to bury the Claude usage. If you find us mid-search and this doesn't match, we will happily recommend someone better.
Read the thinking that shapes us.
The library is as much a part of the company as the services. We think in public and ship what we're ready to defend.
Essays
-
The handbook as thesis, defended38 min / Why the living document replaces the consulting deck.
-
A consulting firm's P&L on Claude19 min / The unit economics behind our pricing.
-
The operator-to-senior ratio12 min / Why we inverted the pyramid.
Hire us for an engagement.
Or come work with us.
We're currently hiring operators for the security and marketing lines. Senior people only, remote, with strong writing chops.