Applied AI / Capability
AI workflow automation.
Use models for one bounded step in a process, such as sorting or drafting, with a person approving the result.
Describe the task
When to bring us in
Sound familiar?
- Requests pile up because someone has to read and route each one by hand.
- Our team writes the same kinds of replies over and over.
- We want automation, but not something that acts without anyone checking.
The cost of leaving it
Requests queue until someone has time to read and route them, so response times depend on how busy the day is. The same replies get written again and again by hand.
Why us
What we bring to ai workflow automation.
A person stays in charge
Models work on bounded tasks inside your systems, and anything that matters goes to a named person to confirm.
Tested on your examples first
We measure a model on real cases from your work before it reaches staff, so you see where it fails while the stakes are low.
We say when AI is the wrong tool
Not every task should be automated. When conventional software or a person is the better answer, we tell you.
What the work includes
Automate one step, and keep a person on the decision.
We pick a step in an existing process that is repetitive, well understood and easy to check, such as routing a request or drafting a reply. The model does that step and nothing else. A person approves anything with consequences, every action is logged, and the process still works when the model is unavailable.
Step selection
Identify the step, its inputs and outputs, and what a wrong result would cost.
Model integration
The model is called for that step only, with structured output the workflow can check.
Approval points
People confirm or correct results before anything is sent, paid or changed.
Logging and fallback
A record of each decision, and a manual route when the model is unsure or unavailable.
Use cases
Projects we take on.
Routing maintenance requests
Incoming requests sorted by trade and urgency. A coordinator confirms each one before a job is dispatched.
Drafting replies to enquiries
Suggested responses to routine customer emails, which staff edit and send themselves.
Summarising notes for handover
Long case notes condensed into a short summary that the person handing over checks before it is shared.
What you get
Outputs you can review.
- Process map with the automated step marked
- Evaluation set from real past cases
- Automated step integrated into the workflow
- Approval screen for reviewers
- Decision log and fallback procedure
Technologies we use for this
Where it matters
- Health & social care Staff can look up care policies and procedures with the source shown; decisions about a person’s care stay with people.
- Property & field services New repair requests, with their descriptions and photos, can be sorted by trade and urgency before an operator confirms dispatch.
- Fintech & professional services Statements and application forms can be read into your systems field by field, and any value below a set confidence goes to a reviewer.
What’s changing
Shifts to plan for.
July 2026
EU high-risk rules move to 2027
The AI Omnibus entered into force on 27 July 2026, moving obligations for Annex III high-risk systems, such as those used in employment, credit scoring and education, to 2 December 2027.
August 2026
EU AI transparency duties apply
Article 50 obligations, such as telling people when they are dealing with an AI system, apply from 2 August 2026.
February 2026
New UK rules on automated decisions
The Data (Use and Access) Act 2025 widened when significant decisions can be automated, provided people are told, can contest the outcome and can ask for a human review.
Process
How the work runs.
Gather past cases
We collect real examples of the step alongside the decision a person made, and that becomes the test set.
Fix the output shape
The model returns structured fields, such as a category and a reason, which the workflow can validate.
Run alongside people
For a period the model suggests and staff decide, so accuracy is measured on live work before it counts.
Widen it carefully
Only categories that prove reliable move to lighter review, while the rest stay with a person.
Cost
What shapes the estimate.
Variety of inputs
Free-text emails and attachments are harder to handle than structured forms.
Cost of a wrong result
Higher-risk steps need more evaluation, fuller review screens and more detailed logging.
How the work arrives
An inbox, a queue and an API each need different plumbing to connect.
Ways to work together
A first, fixed stage
Discovery, a prototype or a technical review with agreed outputs, so the next decision is made with something concrete in hand.
Delivery in phases
The build split into releases, each with a scope you review and approve before it starts.
Ongoing engineering
An agreed scope for support and improvements after launch, with response expectations written down.
Common questions
Before you get in touch.
Can the automation act without anyone approving it?
Only for low-risk steps you choose, after evaluation shows it performs well enough. Anything with financial, legal or care consequences keeps a person in the loop.
Does it work with our existing systems?
It needs a way to receive inputs and return results, such as an API, inbox or queue. We check what your systems expose before committing to a design.
What affects running costs?
Volume, the size of each input, the model chosen and how often results need review. We measure these during evaluation so the trade-offs are visible.