Service 01
AI Systems
AI systems built around how your business actually runs.
We design and implement AI-assisted systems that qualify inquiries, route requests, read documents and retrieve knowledge — connected to the tools your team already uses, with clear rules for when a person takes over.
You need this if
A high-volume part of your business depends on reading, sorting or answering — and people have become the bottleneck.
01Who it is for
Service businesses with steady inbound demand
Clinics, real estate, education, professional and B2B services — anywhere response speed decides whether an inquiry turns into a customer.
Operations teams working through documents
Teams that read, check and re-key the same kinds of forms, invoices, contracts or reports every week.
Organisations with knowledge trapped in files
The answers exist in policies, manuals and past projects, but finding them depends on asking the right person.
02Problems it solves
- 01
Inquiries wait for someone to be free
Leads arrive at night, at weekends and during busy hours. Some go cold before anyone replies.
- 02
Qualification is inconsistent
Different people ask different questions, so sales time goes to prospects that were never a fit while good ones wait.
- 03
Documents are processed by hand
Data is read from PDFs and emails and typed into other systems, with errors that surface weeks later.
- 04
Knowledge depends on who is available
The same internal questions are answered again and again by the few people who know where to look.
03Example use cases
AI-assisted lead qualification
Every inquiry is answered, asked the questions that matter to you, scored and routed — with the reasoning visible.
Customer inquiry routing
Messages from email, forms and chat are classified by intent and urgency and sent to the right person or queue.
Controlled conversational interfaces
Assistants that answer from approved content only, know when to stop, and hand over to a person with context.
Internal business assistants
Help for staff on procedures, policies and product information, limited to what each role is allowed to see.
Document understanding
Extraction and checking of data from invoices, forms, contracts and reports, with low-confidence fields flagged for review.
Knowledge retrieval
Search over your documents that returns answers with their sources, so people can verify before they act.
System integrations
AI steps connected to your CRM, inbox, messaging, databases and internal tools — not a separate app to babysit.
Reporting and operational automation
Summaries and reports assembled from operational data on a schedule, with anomalies called out.
04What is included
Discovery & design
- Process mapping and opportunity assessment
- Data, access and integration review
- Success measures agreed before build
- Failure modes and escalation design
Build
- Model selection, prompting and retrieval design
- Integrations with CRM, inbox, messaging and databases
- Guardrails, fallbacks and human hand-over
- Controls for content, rules and thresholds
Operate
- Evaluation set and acceptance testing
- Logging and monitoring
- Documentation and team handover
- Tuning based on real usage
05How we implement it
Map the workflow
Follow real cases end to end and find where time, quality or revenue is lost.
Define what good looks like
Write the evaluation criteria and the measures the system will be judged on before building it.
Build a controlled first version
Ship a narrow scope connected to live tools, with human review switched on.
Launch with monitoring
Log decisions, track the agreed measures, and review edge cases weekly.
Extend what works
Widen the scope only where the data shows the system is reliable.
Where we draw the line
- AI does not make final decisions with legal, financial or medical consequences without human review.
- If a rule-based workflow solves the problem reliably, we recommend it instead of a model.
- We document what the system cannot do before launch, not after.
- Sensitive data stays with providers and in regions you approve, and only the data a task needs is sent to a model.
06Related work
All projectsPublished case studies for this service are being prepared. Ask about relevant experience in your first conversation — we will be specific about what we have and have not done.
Request an AI Systems Audit07Questions
Which AI models do you use?
We choose per task. Commercial models from the major providers are often the right fit; open models hosted in your own environment can make sense for data-residency or cost reasons. The choice is documented and the system is built so the model can be replaced.
Will an AI system replace our team?
That is not how we design them. The aim is to take repetitive reading, sorting and drafting off people's desks, and to get the cases that need judgement to the right person faster.
How do you handle our data?
We map what data the system touches before building, prefer provider settings that exclude your data from model training, restrict access by role, and avoid sending more information to a model than the task requires.
How do we know it is working?
We agree measures up front — response time, qualification accuracy, hours saved, error rates — and build the logging needed to track them. A system that cannot be measured cannot be improved.
How long does a first version take?
It depends mostly on the number of integrations and the state of your data. We give a specific estimate after discovery, and we prefer to ship a narrow first version and extend it.
Talk to us about AI Systems.
Tell us how the work runs today and what you want to change. We will propose a sensible place to start.