AI implementation
AI for the work your staff still do by hand
We find the manual work AI can take over, build the system, and train your staff to run it. We did it for our own company first: AI prepares our briefings, checks our invoices, sorts our email and records our meetings.
A typical build
8 weeks
- Week 0-2
-
Scope
We map the task, the data and who signs off. A written brief, a fixed price. We map the task, the data, and the review process that decides whether the system can go live. You get a written brief with the scope and the fixed price.
- Week 2-6
-
Build
Built on your real data, in your environment, with security and records staff from week one. We build the system on your real data, in your environment. Your security and records staff take part from the first week.
- Week 6-8
-
Training and hand-off
Your staff are trained, the system is documented, and it runs without us. Your staff are trained, the system is documented, and it runs without us. You do not need a retainer to keep it running.
In use, and owned by you
We use it to run our own company
One agent runs these jobs for us every day.
- A briefing on what needs attention
- Weekdays, 9 am and 1 pm
- Email sorting, so only urgent mail interrupts
- Every 5 minutes
- Action items from meeting recordings
- As each recording finishes
- Contracts, insurance and certificates checked for expiry
- Weekdays, 7 am
- Purchase orders checked against spending
- Weekdays, 7 am
- New federal awards checked against our sales pipeline
- Weekdays, 8 am
- Contractor invoices checked against logged time
- Daily, noon
- Client invoices drafted for approval
- 1st of each month
Our first version sent too many alerts
We started with a separate agent for each specialty, and each one decided for itself when to interrupt us. Now one agent makes that decision, and the others became scheduled jobs. We learned that on our own time, so you do not pay for the lesson.
Your security team will ask about permissions
Ours did. Our contractors use the same agent, but it works with each contractor's own permissions and sees only their records. Every client asks about this, and we solved it for ourselves first. See how we run our company on AI.
You can read one of our builds. The toolkit we made for a tax accountant in India is open source, with a write-up of what it cost and why we designed it that way. Read the write-up.
What an AI system should do
We look for skilled people who spend their week on data entry, follow-up and reformatting. The software does that part, shows its sources, and asks before any decision that matters. People keep the judgment.
- A proposal advisor for a National Science Foundation program
- The EPIIC Assistant helps researchers write proposals for the Enabling Partnerships to Increase Innovation Capacity (EPIIC) program. It gives stored expert answers to common questions, and AI answers based on the solicitation and the program guidance for everything else.
- EPIIC program · In use
- An instrument finder that cites every answer
- We built it to answer which facility can characterize a given material, using only the program's own pages. Every claim links to its source, and each answer ends with the named manager of that instrument.
- Materials Innovation Platforms
- The scheduled jobs that run our company
- The briefings, the meeting action items, the expiry checks and the invoice checks from the list above, all running on our own operations. This is where we learned what breaks, before we built any of it for a client.
- Our own company · Since 2025
Two fixed-price ways to start
Each review has a price agreed before it starts, takes two weeks, and ends with a written report. Neither one commits you to further work, and neither fee is credited against a build. If neither fits, book a 30-minute call and we will find the right first step together.
If you have manual work and no plan yet
AI Workflow Review
- You bring the tasks
- Two or three tasks you think AI could take over: the month-end close, the request queue, the annual report, the work your best person spends Thursdays on. We look only at the tasks you bring.
- What you get
- For each task: what it would take to build, what it would save, what could block it, and which one to do first. We also tell you which tasks not to automate, and why. Written for the person who holds the budget.
- Duration
- Two weeks
- Fee
- $12,000 fixed
- Covers
- Up to three tasks
- Your time
- About six hours
- Obligation
- None
If you have an AI project that is stuck
Production Readiness Review
- What we do
- We read the code, talk to the people who built it and the people who must approve it, and check your project against the review it will face.
- What you get
- A written recommendation: launch it, fix it, or stop. We name each issue that will fail review, what it takes to fix, and what finishing the project would cost.
- The fee is not credited against a build
- A firm that credits the review fee against a later project has a reason to recommend that project. Launch, fix and stop are all real outcomes, and stop can save you the most.
- Duration
- Two weeks
- Fee
- $12,000 fixed
- Covers
- One project or system
- Your time
- About four hours
- Obligation
- None
We reply within two business days.
Bring us a task you think AI could take over.
In a 30-minute call, we will tell you if it is worth building, what it would cost, and how long it would take.