- Home
- AI development
AI development
AI development, built into the work your team already does
We build AI agents, private assistants over your documents, voice agents and the integrations that tie them to your CRM, inbox and calendar. Each project starts with one task and a published price, and you get a first working version two weeks after we have access.
- Prices published on the site
- First working version in two weeks
- Support month to month, cancel any time
From one task to a system your team uses every day
We plan for daily use from the first week: who approves what the system does, where the data sits, and what it costs to run each month.
Scope
A written audit of one task: what the system will do, what it needs access to, which band it lands in.
Build
We build in two-week steps. The first working version runs on your real data, behind a login.
Ship
Your team uses it with approvals switched on. Payments, deletions and anything sent to a customer stay with a person.
Operate
On a month-to-month support plan we host, monitor and update it, including when a model version is retired.
What we build
- AI agents Agents that finish a task, with your sign-off
- Generative AI Drafts, summaries and documents from your data
- Claude development Apps and agents built on Anthropic's Claude
- MCP servers Connect AI assistants to your own systems
- Private AI over your files Answers from your own files, on your server
- Semantic search Find the right document by meaning and keyword
- Language processing (NLP) Sort, extract and summarize text
- Computer vision Read photos and scanned documents
- Multimodal AI Text, images and voice in one system
- AI for calls and chat Answers calls and books the appointment
- LLM fine-tuning Train a model on your own examples
- LLM evaluation Test which model does your job best
- MLOps Run, monitor and update models in production
- Expert systems Your rules, applied the same way every time
- Reinforcement learning Decisions learned in a simulator first
- Local AI setup AI on hardware you own, set up on-site
- AI workflow automation Documents, approvals and multi-step processes
- AI for small business One task automated first, priced up front
- Enterprise AI Security review, procurement, your cloud
- AI readiness assessment A written review of where AI fits
Common tasks and the service for each
Start from the task you want off your team's plate. These are common starting points and the band they usually land in; the free written review names the band for your case.
| The task | What we build | Usually lands in |
|---|---|---|
| Answer calls, book the appointment, send a summary | AI for calls and chat | One process, $2,500 to $6,000 |
| Turn emailed invoices into draft bills for review | A finance agent | One process, $2,500 to $6,000 |
| Let staff ask questions of thousands of pages of policies | Private AI over your files | Connected systems, $6,000 to $15,000 |
| Read photos of paper forms into a database | Computer vision | One process, $2,500 to $6,000, or connected systems with integrations |
| Let Claude or ChatGPT look up orders in your own system | An MCP server | One process, $2,500 to $6,000 |
| Make a smaller model write in your house format | Fine-tuning | Connected systems, $6,000 to $15,000 |
| Several systems, roles and an AI layer across them | Custom software | Full build, from $15,000 |
AI agents for the teams that carry the paperwork
An agent reads what comes in, looks things up in your systems and prepares the next step. A person approves anything that spends money, and sends anything addressed to a customer. Each department page lists the agents we build for it, and What AI agents do has the full task list.
- AI agents for finance Invoices, reconciliation and month-end close
- AI agents for HR Onboarding, policy answers and paperwork
- AI agents for legal Contract review, intake and deadlines
- AI agents for operations Scheduling, vendors and daily reports
- AI agents for procurement Quotes, purchase orders, invoice matching
- AI customer support Chat that answers and hands off to a person
Three published bands, decided by scope
One workflow running end to end, or one internal tool.
Several systems made to work as one: integrations, roles, reports.
Custom software built from scratch, over a longer schedule.
The band depends on how many systems the AI has to touch and how many roles and steps it covers; the model underneath changes it little. A single agent on one task is usually one process. A private assistant over several document sources, or an agent working across three systems, is usually connected systems.
Care $300 to $500 a month. Active $750 to $1,500 a month. AI model usage, GPU time and hardware are paid to the provider, on accounts in your name.
What people ask before an AI project
What is the difference between AI development and AI automation?
AI automation connects tools you already have, for example a phone line, a web form, a CRM and a calendar, with AI handling one step in the chain. AI development builds the AI part itself: an agent, a search over your documents, a fine-tuned model, an evaluation suite. Many projects need both, and we price them together.
Which AI models do you build with?
We build with Claude or Gemini, or with an open-weight model such as Llama or Qwen when the data has to stay on your own server. If your company already pays for OpenAI, we can build on it. We pick the model per task, test it on your examples, and write down which model the system calls.
Where does our data go?
The application, database and document index can live on your own server or cloud account. If the system calls an outside AI model, we tell you which one and what data it sees before we build, and the account with that provider is opened in your company's name.
How long does an AI project take?
The first working version is ready two weeks after we get access. One process takes 2 to 4 weeks in total, connected systems 4 to 8 weeks, and a full build from 8 weeks. The clock starts when the accounts and sample data are in hand.
Do we need clean data before we start?
No. The audit looks at what you have: spreadsheets, PDFs, email, a CRM export. If something has to be cleaned or merged first, that work is part of the scope and the price, and we say so in writing before you sign.
Who owns the system when it is done?
Accounts and code repositories are opened in your company's name from the first day. The custom code we write is assigned to you under the contract once it is paid for; open-source libraries stay with their owners under their licenses, listed at handover. Support is month to month, and you can take the system to another team at any time.
Also on this site
Tell us what's slowing you down.
You get a free written review by email within two business days.