Services
Everything it takes to get AI working
Six services covering the whole path, from finding the work worth automating to keeping it running a year later. The first one is the recommended place to start.
Service 01
AI opportunity audit
Before anything is built, we work out what is worth building. We sit with the people doing the work, follow a job end to end, and count what it really costs in hours, delays and mistakes. Then we rank the candidates by payoff against effort.
You finish with a document you could hand to any competent team, including one that is not us.
Start with an auditWhat you get
- A ranked list of automation candidates, each with an estimate of hours saved
- A plan in priority order, with what each project takes to build, what it takes to run, and what it should give back
- A map of the systems involved and what access each one needs
- An honest note on anything we think is not worth automating yet
- A walkthrough session with your team, not just a document drop
Service 02
Custom AI agents
An agent is software that can read your systems, decide what to do, and do it, within limits you set. We build them one job at a time, because a system with one clear job can be tested, measured and trusted.
- Scheduled work: the Monday report, the month-end pack, the daily reconciliation
- Triggered work: a new ticket arrives, an order fails, a threshold is crossed
- On-demand work: someone asks a question in plain language and gets an answer with sources
Built in from day one
- Approval gates on anything that spends money, contacts a customer or changes a record
- A full log of every run: what it read, what it decided, what it changed
- Fallbacks: when the agent is not confident, it escalates instead of guessing
- Cost ceilings, so a runaway process cannot quietly run up a bill
Service 03
Systems integration
AI is only useful if it can reach your data and act in your tools. Cevran connects to 8,000+ business applications, so the output lands where the work continues.
Where your team talks
Slack, Microsoft Teams, email, Telegram, Discord. Agents post, answer and ask for approval in the channel people already watch.
Where your records live
CRMs, help desks, accounting packages, ERPs, project trackers, spreadsheets, databases and storefronts.
Where your documents sit
Google Drive, SharePoint, Notion, Confluence, shared drives and the PDF archive nobody has opened in years.
If something you run is unusual or built in-house, it can usually still be connected. Bring your list to the call and you will get a straight yes or no on each system.
Service 04
Your data, made usable
The most common cause of a disappointing AI project is not the model. It is the state of the information underneath it. If that information is scattered, contradictory or locked in scanned PDFs, an assistant will answer confidently and wrongly.
We prepare the ground: gather the sources that matter, resolve the contradictions with someone who knows the answer, and structure it so the AI retrieves the right passage and cites it.
What this involves
- Finding the authoritative version of each document, and retiring the rest
- Making scans and PDFs machine-readable
- Answers with citations, so a person can check the source in one click
- Permissions that mirror your own, so the assistant is scoped to what each person can already open, and we document exactly how that mapping works for every system we connect
Service 05
Rollout and training
A system nobody trusts gets bypassed within a month. We introduce it properly: what it does, what it will not do, how to check it, and who to tell when it is wrong.
- Hands-on sessions with the people who will use it daily
- Written guidance in your language, not ours
- A named path for reporting problems, and a habit of fixing them fast
- A staged rollout: one team first, then wider once it has earned it
Why adoption fails
Adoption failures are rarely technical. They happen because nobody explained where the output comes from, so the first surprising answer destroys confidence permanently.
The fix is unglamorous: show your work, cite sources, and make it easy to correct. We build for that from the start.
Service 06
Managed operation
Software that touches five other systems needs someone watching it. APIs change, formats drift, and the business six months from now is not the business the system was built for.
Monitoring
Every run is recorded. Failures raise an alert to us, not a silent gap in your Monday report.
Maintenance
When a vendor changes an API or a model is retired, we handle the migration. That is our problem, not yours.
Improvement
We review what the system did, where it handed back to a person, and what it should learn next, then extend it.
Engagements
Three ways to work with us
Fixed scope and a fixed price, agreed in writing before anything starts.
| Engagement | Best when | You get | Commitment |
|---|---|---|---|
| Audit | You know AI should be doing something here, but not what. | A prioritised plan you own outright. | Fixed fee, about two weeks. |
| Build | The target is clear and you want it working. | A working system in your tools, tested and handed over. | Fixed scope, fixed price, per project. |
| Run | It is live and you would rather not staff it yourself. | Monitoring, fixes, improvements and a named contact. | Monthly. The notice period is agreed in writing before you start. |
Pricing depends on how many systems the work touches and how much you want us to run. We put a fixed number in writing after the call.
Not sure which one you need?
That is what the first call is for. Bring the process that costs you the most time.