Loading...
Human-in-the-loop AI, UK
The AI agency UK teams trust, with a person in the loop
We are an AI agency UK organisations use to build AI that does the volume, while their people keep the judgement.
Built for →
Live routingHover to pause
Incoming
Checking…
AI handles0
A person decides0
UK firms (10+ staff) using AI0%Up from around 12% in late 2023.
Source: Office for National Statistics ↗Large UK firms (250+) using AI0%Against 28% of firms with 0 to 9 staff.
Source: Office for National Statistics ↗What that means
Most firms now use AI. Few have decided who checks it.
What we do
An AI agency UK firms can hold to account
Most businesses do not need another strategy deck about AI.
They need a few working systems that save real hours, and someone in the business who owns them.
As an AI agency UK teams bring in to get that done, we build AI that knows when to ask a person.
The AI reads, drafts, sorts and extracts at speed.
A named person approves anything uncertain, sensitive or significant, with the full context in front of them.
That line matters in law too, because the ICO expects meaningful human review, not a rubber stamp.
AI drafts
Check
Person decides
Learns
01 / 04
Services
Three ways we help
Consultancy and automationScoping, builds and implementation, with the approval step designed in.
Training for people who review AIFor leaders, whole workforces and governance teams.
Governance and assuranceReadiness, policy, assurance and front-desk AI with safe handover.
Sectors
AI for regulated work, where someone signs off
The professions with the most to gain are the ones where a qualified person must still own the decision.
SectorsAI for accountants →AI for accountants, with a qualified person still signing offSectorsAI for law firms →AI for law firms, with a solicitor checking every outputSectorsAI for financial advisers →AI for financial advisers, where the adviser still owns the adviceSectorsAI for mortgage brokers →AI for mortgage brokers, with the adviser owning the recommendationSectorsAI for estate agents →AI for estate agents, with a person checking what mattersSectorsAI for HR →AI for HR, with decisions about people left to peopleSectorsAI for recruitment →AI for recruitment, with a recruiter behind every noSectorsAI for architects →AI for architects, with design authority kept in the studioSectorsAI for charities →AI for charities, with trustees still accountableSectorsAI for dentists →AI in dentistry, with the dentist still making the call
Where the line sits
You set the bar. We build to it.
Every system we build has a confidence bar that decides when the AI may act alone.
Raise it and more work comes to a person.
Lower it and the AI handles more on its own.
Some decisions go to a person whatever the score, such as anything with a legal or similarly significant effect on someone.
Try the dial to see how the split changes.
What the AI handles
- First drafts of routine emails and reports
- Extracting data from documents
- Sorting and routing inbound work
- Summarising long threads and calls
What stays with a person
- Decisions with significant effects on people
- Anything that commits the business
- Judgement calls the rules do not cover
- Complaints and challenges to AI output
85%
AI acts alone
130
65% of 200 cases
Goes to a person
70
35% of 200 cases
Illustrative batch with a typical confidence spread. Your real split comes from a pilot on your own work. Some decisions go to a person whatever the score.
Live demo
Talk to it. Then ask for a person.
This site runs its own assistant, by chat in the corner and by voice here once you sign in.
Ask it about our services, or ask it how the handover to a person works.
It is the same pattern we build for clients, running in front of you.
Voice Demo
Loading...
01
Map
Find the hours
We map where time goes and pick one or two workflows where AI can take volume without touching judgement.
02
Draw the line
Decide what needs a person
Together we set the rules, thresholds and the named reviewer for every output that matters.
03
Build
Ship it with your team
We build into the tools you already use, with logs, escalation and a stop switch from day one.
04
Hand over
Train and let go
Your people learn to review AI output properly, and the system improves from every correction they make.
Questions
Asked and answered
Further reading
From the source
Information Commissioner’s OfficeHow do we ensure individual rights in our AI systems? ↗Why a rubber-stamp review is not meaningful human oversight.European Commission AI Act Service DeskEU AI Act, Article 14: Human oversight ↗High-risk AI must let people monitor, override and stop it.Government Digital ServiceAI Playbook for the UK Government ↗Principle 4: meaningful human control at the right stages.Department for Science, Innovation and TechnologyIntroduction to AI assurance ↗The UK government’s primer on assuring AI systems.Office for National StatisticsArtificial intelligence in UK businesses: 2023 to 2026 ↗How many UK firms use AI, by size.
Next step
Tell us the work. We’ll show you where the line goes.
A 30-minute call. We map one workflow, what the AI could take, and where a person must stay in the loop.