What we do · AI-native software & agents
AI-native software for accountancy & tax firms.
We build AI tools and automation that take repetitive work out of accountancy and tax practices, and we run them on the IT environment we already secure. That combination is the point: an AI tool is only as trustworthy as the environment it runs in and the people accountable for it.
What we build
Tools for the gaps between your systems.
Your practice software does its core job well. The time goes on everything around it: the chasing, the re-keying and the hand-offs no product covers.
- Files
- Users
- Settings
- Deleted
- Notifications
- Passkeys
Example Trading Ltd
File browser
- Clients
- Onboarding4
- Engagement letters
- Templates
Northfield Joinery Ltd
New client · Limited company · Started 2nd Nov 2026
- Engagement letterSigned by Dan Evans3rd NovDone
- ID and AML checksBoth directors verified3rd NovDone
- Client portalDan Evans has signed in with a passkey4th NovDone
- HMRC agent authorisationRequest sent; reminder sent on 5th NovWaiting on client
- First records2 of 5 receivedIn progress
- Last year’s accounts
- Bank statements
- VAT registration· requested
- Payroll details· requested
- Loan agreements· requested
- Incoming5
- Filed
- To check1
- Rules
Incoming documents
Read, labelled and filed as they arrive
23 filed today1 to check
letter.pdf needs a person
It looks like a letter from HMRC, but the reference on it doesn’t match Maria Clarke’s records, so it hasn’t been filed.
Open itChoose where it goes
- Automations12
- Runs
- Connections
- Settings
When a client uploads a document
38 runs this week
- When · Client portalA client uploads a document
- Then · Document intakeRead it and label it by type and tax year
- Then · Practice softwareAttach it to the client’s job and move the job on
- Then · Microsoft TeamsTell the preparer
- Home
- Search
- Policies
- Procedures
- People
- News
Search
Policies, procedures and people
- Expenses policyPolicyWhat you can claim, the limits, and the receipts to keep.Version 3 · Updated 2nd Oct 2026
- Claiming mileageProcedureLog business journeys in the expenses tool and submit them at the end of each month.Finance · Updated 14th Sep 2026
- Paying filing fees for a clientProcedurePay with the firm card, then recharge the fee to the client’s job.Finance · Updated 3rd Jun 2026
- GOGrace OkaforPersonPractice manager. Approves expenses and travel.Operations · Extension 214
- Timesheet
- Clients and jobs
- Invoices
- Reports
Timesheet
Sam Patel · Week of 2nd November 2026
- Dashboard
- Jobs
- Deadlines
- Preparers
Self assessment 2025-26
640 returns · Deadline 31st January 2027
Filed
412
+38 this week
Ready to review
37
+12 this week
In progress
95
−29 this week
Waiting on records
96
−21 this week
Filed, by preparer
- Sam Patel112 of 150
- Jo Bennett96 of 140
- Chris Lowe84 of 130
- Amira Haddad70 of 120
- Tom Reid50 of 100
- Approvals3
- Activity
- Agents
- Settings
Approvals
Nothing an agent writes goes to a client until a person approves it
Hi Sam, could you remind me what I need to pay in January?
Draft reply
Hi Rhys,
For January you’ll need to pay £2,140 by 31st January 2027. That’s a balancing payment of £860 for 2025-26 and your first payment on account of £1,280 towards 2026-27. Pay HMRC online with your UTR followed by the letter K as the reference.
Sam
From:2025-26 tax computationPayments on account
Approve and sendEditReject
Built to work together.
Every tool we build connects to the others and has its own API, so the steps between them can be automated, by your other systems or by AI agents, instead of re-keyed by your team.
Client data
AI you can trust with client data.
Most firms’ first worry about AI is client data, and that’s the right place to start. AI can produce a working tool in days, but it won’t make it secure on its own. We build to the same standard as the IT we secure, and run what we build, so one partner is accountable for both.
It’s checked for security, not just for working
AI writes code fast but not safely by default. Everything we build is real code in a proper repository, reviewed and tested for security before release. No low-code assembly.
Your data stays consistent
Tools work from your practice software as the master record, rather than keeping copies that drift out of step with it.
It signs in with the accounts you already use
Tools use your existing sign-in, so your MFA and access rules apply, and disabling a leaver’s account removes their access everywhere.
It only reaches what it needs
Each tool gets access to the data its job requires, and nothing else.
You know where client data goes
For every tool, we can tell you which services process client data, where, and on what terms.
Someone is accountable after launch
We host, patch, monitor and support what we build, for as long as you use it. Nothing gets handed over and forgotten.
How we build
Small, working, then run properly.
- 1
Start with the problem
Usually one we’ve spotted while running your IT, or one your Technology Review surfaced. We start from what your team actually does.
- 2
Ship a small first version
AI has made building fast, so you see something working in your own setup early, then we improve it with the people who use it.
- 3
Run it
Once it’s earning its keep, we host it, secure it and keep it working as your systems change, and it’s built so the next feature doesn’t mean starting again.
“The product exceeded expectations in both quality and functionality. The JDLT team continues to work quickly and efficiently.”
Case study · A UK tax practice of about a dozen people
A client portal the practice owns, in place of ShareFile
A UK-hosted client portal for a small tax practice moving off ShareFile: passkey sign-in, its own branding and storage, and a record of every action.
Read the case studyFurther reading: custom software vs off-the-shelf and building business software: 18 keys to success.
Questions
Common questions
Is it safe to use AI with client data?
It can be, on the right terms. Use business versions of AI tools that don’t train on your data, under a data processing agreement, set up so each tool reaches only the data it needs, with a person checking anything that goes to a client. Free and personal accounts aren’t the place for client data.
Can accountants use ChatGPT, Copilot or Claude with client data?Do we need to be a managed IT client?
No, but it’s where we’re most useful. When we already run your environment, we know your systems and data before we start, and the tool lives somewhere we’re responsible for.
Will it replace our practice software?
No. CCH, IRIS, TaxCalc and Xero do their jobs well. We build in the gaps between them and leave your practice software as the master record.
Do you offer AI strategy or training?
AI strategy is part of the conversation in a Technology Review rather than a separate deliverable. What we sell is working software, built and run for you.
What does it run on?
Full-stack TypeScript, hosted on AWS or in the cloud you already use, and signing in with your existing accounts.
How is it priced?
Usually by the hour, which suits building in small steps and changing course as you learn. Where the scope is well defined and stable, we quote a fixed fee.
Where to start
Start with a Technology Review.
A short, paid review of your IT, your security and where AI could take real work off your team. It’s the first step with every new client, and it shows you what’s worth doing before you commit to anything bigger.
- Half a day to a day
- A written report you keep
- Price agreed before we start
- 1
Review
You start hereWe look at your setup, your security and the workflows that eat your team’s time.
- 2
Secure
We fix the foundations, so anything we build sits on a setup you can trust.
- 3
Automate
We build and run the tools that make the biggest difference to your practice.