Managed AI operations for UK accountancy practices

A week of client chasing. Quietly handled.

Mach Lilies installs and runs a human-controlled helper that finds missing records, drafts chasers, classifies replies, attempts agreed filing and tracker writes under separate rules, and routes judgement calls to your team.

Every client-facing message starts with a person’s approval. Professional judgement stays with your qualified team.

£799/month, no setup fee, excluding VAT · first month after go-live free and measured

Live client deployments

Proof, not promises.

Mach Lilies AI Helpers are already supporting real finance teams inside the inboxes where repetitive query and payment work happens.

1-800Accountant

  • Ongoing live pilot
  • Live since

From unread query to review-ready response.

  • 67% less Reported time spent Internal pre/post staff comparison
  • $666K Estimated annualised gross staff-capacity value Client labour-hour model

Client-reported internal pre/post comparison and labour-cost assumptions. The $666K figure is an estimated annualised gross staff-capacity value, not an independently audited cash saving. Ongoing proof-of-concept deployment.

The Helpers prepare drafts; people retain control of client-facing decisions.

See the 1-800Accountant story →

inDinero

  • Ongoing live pilot
  • Live since

Two AI Helpers coordinate the payment chase.

  • 47% less Human time spent Across the measured workflow
  • 82% shorter End-to-end payment cycle Across the full payment loop

Client-reported internal comparison during the ongoing deployment. The 1,780 figure describes the reported global agent network, not the number of completed payment cases. Results are not independently audited.

Financial validation, exceptions and payment decisions remain with people.

See the inDinero story →

These deployments are not presented as MTD implementations. They prove the operating pattern behind Quiet Quarter — monitor the inbox, identify what is missing, prepare the next action and stop for human judgement — now packaged for UK accountancy practices.

See both deployments in full, with the methodology behind each claim →

01 · The working week

One client. One record. Monday to Friday.

The same designed sample moves through the practice’s tracker, approval route, inbox and folders. It is a front-end simulation—not a customer case study, live integration, or proof of delivered results.

The working week · three chapters

Simulation · sample data

Case QQ-SIM-042 · Priya Sharma · April–June 2026 records

  1. 01Monday · 08:47–09:02

    Find and draft

    The agreed tracker — not a new dashboard — shows Priya Sharma still owes the June bank statement. A chase is drafted from the approved template, in the practice’s tone, and held. Nothing sends.

    Missing record identifiedChase drafted · held

  2. 02Monday 09:04 – Tuesday 14:32

    Approve, receive and update

    A person approves the chase and it leaves from the practice address. The client replies with the statement; it is filed in the agreed folder and the same tracker row moves from 2 outstanding to 1.

    Approved by a personReply classified · filedTracker 2 → 1

  3. 03Tuesday 14:33 – Friday 17:02

    Stop and summarise

    “Is the van repair an allowable expense?” stops the workflow: the question routes to Kelly · client team with context, and no answer is sent. Friday’s summary shows what moved, what remains outstanding and what needs a human.

    Judgement routed to a personFriday summary produced

Inspect the complete Monday-to-Friday workflow — all nine stages →

02 · The offer

One offer. The full operating setup.

The full MTD chasing setup: more workflows, more capacity, priority setup, and stronger operating controls.

03 · The boundary

What moves. What stays human. Where it lives.

One controlled path across the tools the practice already runs — with the human checkpoints fixed before anything is built.

another AI platform — in plain English: a managed workflow scoped around familiar tools

01Agreed trackerfind what is missing
02Practice inboxdraft, approve, receive
03The helperclassify and prepare, under agreed rules
04Folder and trackerfiled where agreed · row updated
05Named staff ownerapprovals and exceptions

Moves through the workflow

Chasing, reply classification, agreed filing and tracker updates, exception flagging and the Friday summary.

Stays with your team

Professional judgement, tax and accounting decisions, sensitive replies, and every approval at launch.

Where it lives

Your existing inbox, tracker and document folders — exact access and any integration confirmed before build.

Client-facing send
Held for a person at launch
Professional judgement
Stays with your qualified staff
Sensitive or unclear reply
Routed to a named owner
Access and processing terms
Agreed before connection

The data-safety page sets out the proposed control framework — access, write rules, action-record coverage, escalation and recovery. Inspect the control model →

04 · Fit and rollout

A good workflow has a clear start, finish and human owner.

The fit call is allowed to end with “not yet”. The scorecard gives you the lower-commitment version first.

A strong fit

  • A UK accountancy or bookkeeping practice
  • Recurring, email-led client-record chasing
  • A tracker or source of truth your team actually trusts
  • Enough weekly chasing volume for released time to matter

Not a fit (yet)

  • Chasing runs mainly on phone, paper or WhatsApp
  • No reliable tracker and no named workflow owner
  • Very low recurring chase volume

The scorecard applies the same versioned rule set as the full fit criteria — exact bands, thresholds and edge cases included.

  1. Fit callan honest go or no-go
  2. Map the ruleswho is chased, who approves
  3. Approval-mode pilotnothing sends unapproved
  4. Measure, review, widenoutputs against the signed plan

See the full implementation path and the first month live →

Book a fit callTake the readiness scorecard

05 · Questions

What a careful practice asks before buying.

Does the helper give tax or accounting advice?

No. It handles the chasing, filing, tracking and reporting around client records. Judgement — tax treatment, accounting decisions, anything regulated — stays with the qualified people at your practice.

What does "human approval mode" mean?

At launch, every client-facing message is held for approval by your team. That remains the rule throughout the live approval-mode pilot. Only after that pilot is reviewed may the practice consider a workflow-specific, tested, written change for later operation.

Do we need to change software?

Not necessarily. The starting point is your existing inbox, tracker, and document locations. The fit call confirms whether they can support a controlled workflow and what access, integration, or extra tooling would be required. No software change is assumed or promised before that review.

How quickly can we be live?

For planning, allow at least 21 working days after access and scope sign-off, including a live approval-mode pilot. This is a planning allowance, not a promised delivery date; the signed proposal confirms the actual schedule. Delayed access, unsupported systems, or scope changes can extend it.

More questions — offer mechanics, acceptance reviews, edge cases and data handling — are answered on Pricing and Data safety.

The next useful step

Bring us one chase loop.

We will tell you whether it is worth installing, what can move safely, and what must stay human-led.

Book a fit callTake the readiness scorecard