Real work. A clear roadmap.
Last updated: 7 November 2026
26 roles. 351 tasks. One roadmap.
80%
Share of the work AI can support today (280 of 351 tasks)
71
Tasks that stay fully human (20%)
24
Tasks we built first
3
Delivery waves
People moved to the work that needs them: coaching, escalations and financial advice to the business.
Augmentation % is the share of a role's tasks that AI can support today.
Before
AI was being tested team by team, with no shared view of where it mattered. Finance closed the month with spreadsheets and manual prep. Customer service quality was checked by manual sampling.
What Measure identified
Modern Milkman: 26 roles and 351 tasks mapped. We looked at two layers.
1. What can AI do? AI can support 280 of the 351 tasks (80%). 71 tasks stay fully human (20%).
2. What can AI do with the systems in place today? That is where the difference lies. 24 tasks could start straight away, so we built those first. 97 tasks were blocked by internal systems: access, integrations and the database. We then helped get those systems in order, so these tasks can be automated too.
The roadmap runs in 3 delivery waves.
Finance: before / after
87%
Finance Manager
Share of the work AI can support today
86%
Accounting & Operations Manager
Share of the work AI can support today
Before
Spreadsheets and manual prep at month-end.
After
A Finance Close Agent over Xero that drafts management-account commentary, assembles audit working papers, prepares consolidations and answers colleagues' finance questions. People review every output.
ROI measured on
- Working days to close
- Analyst hours on manual prep
- First-pass acceptance of agent drafts
- Time moved to business partnering
Customer Service: before / after
67%
Customer Success Ambassador
Share of the work AI can support today
AI drafts replies across chat, email and social, looks up orders and delivery issues, and turns QA feedback into coaching. Escalations and complaints stay with people.
58%
Customer Success Team Lead
Share of the work AI can support today
Reporting, QA and performance monitoring move to AI. Leading the team, 1:1s and escalations stay human.
Before
Quality checked by manual sampling.
After
An agent over Dixa and Looker that gives a live team-performance summary with exceptions and moves QA from manual samples to full coverage. Coaching stays with people.
ROI measured on
- QA coverage
- CSAT and response time
- First-contact resolution
What ROI per function looks like.
One finance function. Seven workstreams. Every one scored, built and measured.
Month-end close
Close agent that prepares journals, reconciliations and checklists
- Augmentation score
- 86%
- Hours saved / month
- 72
Management commentary
Agent that drafts monthly commentary from the ledger
- Augmentation score
- 92%
- Hours saved / month
- 28
Consolidation
Multi-currency consolidation agent
- Augmentation score
- 84%
- Hours saved / month
- 30
Audit preparation
Audit pack assembler for working papers
- Augmentation score
- 88%
- Hours saved / month
- 25
Invoice processing
Invoice agent: capture, match and route for approval
- Augmentation score
- 94%
- Hours saved / month
- 64
Finance questions
Self-service finance assistant for colleagues
- Augmentation score
- 90%
- Hours saved / month
- 30
Cash forecasting
Forecast co-pilot, with people making the call
- Augmentation score
- 71%
- Hours saved / month
- 18
86%
Average augmentation score
267 hours
Saved per month (3,204 per year)
€208,000
Capacity per year
€72,000
Investment in year 1 (build + run)
2.9x
Return in year 1
4 months
Payback
Freed-up time goes back to better work: advice, customers and decisions.
Illustrative model based on a typical 6-person finance team at €65 per hour. Your report uses your own roles, hours and costs.