Planned and reactive maintenance of heating, ventilation, electrical and fire systems for schools, care homes, offices, hospitals and light industrial sites across the West Midlands and Worcestershire. 96 staff; three people hold the key accounts.
Verdict: The programme covers 24 accounts still customers (of 25 in the reviewed group a year ago): 30 reviews fall in the next six months (183 hours), and 4 due within 30 days; 4 with a pack started (3 ready, 2 draft in all). 5 contracts end before a review can be held in time. Main measures now: two or more contacts engaged at 11 of 24 in the programme (46%, give or take 19 points); comparison 4 of 14 (29%); renewals reviewed before the hold-by date: reviewed group 1 of 8 (13%); comparison 0 of 3 (too few for a rate); both service levels met: reviewed group 11 of 22 (50%) of those with both levels on file; comparison 5 of 8 (63%); 2 reactive-only accounts to move onto a planned contract; next steps done on time: 2 of 4 (too few for a rate). Spend kept is 98% (give or take 10 points), against 102% for other existing customers: a direction only. Read the main measures again on 16 Jan 2027 and all of them on 16 Oct 2027, the same way. The first year costs £17,275 in fee, set-up and review time. Not yet measured: fill in the Value Case after 16 Oct 2027.
From the Contacts log, by role and the last date someone spoke to or met them. Red: one contact or none, so if that person leaves the account may follow. The comparison group is the 15 accounts left out by alternating down last year's sales list below the ten largest, which are all reviewed.
| Account manager | Oct | Nov | Dec | Jan | Feb | Mar | Reviews | Hours needed | Hours available |
|---|---|---|---|---|---|---|---|---|---|
| Nadia R | · | · | 3 | · | 2 | 3 | 8 | 52 | 108 |
| Tom H | · | 4 | 2 | 1 | 3 | 2 | 12 | 71 | 108 |
| Gareth P | · | 3 | · | 2 | 4 | 1 | 10 | 61 over | 48 |
| Total | 0 | 7 | 5 | 3 | 9 | 6 | 30 | 183 | 264 |
Hours per review are built from who attends and for how long: 8.5 for Tier A (two people), 4.75 for B, 4.25 for C, including preparation, travel and follow-up. Gareth P is over capacity: move an account or send one person instead of two. Reviews are separate from the monthly contract meetings 5 of the largest accounts already have: those cover the jobs list; the review covers the year with the people above it. 5 contracts end before their first review in the calendar can be held in time (210 days ahead for a likely tender, 90 for a direct renewal):
| Account | Review | Status | Sections still missing | Next steps |
|---|---|---|---|---|
| Academy trust (11 schools) | 2 Sep 2026 | Held | nothing | 4 |
| Private hospital | 23 Sep 2026 | Held | nothing | 3 |
| Care home group (8 homes) | 4 Nov 2026 | Ready | nothing | 0 |
| University faculty buildings | 5 Nov 2026 | Ready | nothing | 0 |
| Distribution warehouse operator (3 sites) | 10 Nov 2026 | Ready | nothing | 0 |
| Office landlord (5 buildings) | 12 Nov 2026 | Draft | Value to them; Evidence for the value | 0 |
| Food manufacturer (2 factories) | 26 Nov 2026 | Draft | What we delivered; Service measures (none on Accounts); Value to them; Evidence for the value | 0 |
Each pack pulls its service, sales (planned and reactive), contract and contacts lines from the workbook; the account manager writes the value to the customer in the customer’s terms, with the evidence. “TBC”, “n/a” or a dash count as missing. The two held reviews agreed 7 next steps; 2 of 4 due so far were done by their due date, and 1 is past due (the theatre ventilation proposal).
| Main measure (can move within 90 days) | Reviewed group | Comparison group |
|---|---|---|
| Accounts with two or more contacts engaged | 11 of 24 (46%) ± 19 | 4 of 14 (29%) |
| Next steps done by their due date | 2 of 4 (too few for a rate) | – |
| Renewals reviewed before the hold-by date (last 12 months) | 1 of 8 (13%) | 0 of 3 (too few for a rate) |
| Days from contract end to signed renewal (median) | -7 | +4 |
| Accounts meeting both agreed service levels | 11 of 22 (50%) | 5 of 8 (63%) |
| Reactive-only accounts to move onto a planned contract | 2 | – |
Counts per account, taken the same way each time from the Contacts log, Actions and Accounts. With 24 accounts each is a small number, so read the counts: two or more contacts is give or take 19 points. Next steps start with the first reviews and are too few to rate yet. 12 renewals fall due before the re-measure; 7 of the 12 in the next 12 months have a review in the calendar before the hold-by date.
| Direction only | Reviewed | Comparison | Other existing customers |
|---|---|---|---|
| Spend kept, all sales | 98% ± 10 | 96% ± 9 | 102% |
| of which contract value (planned) | 95% | – | – |
| of which reactive | 107% | – | – |
| Accounts kept | 24 of 25 | 14 of 15 | – |
Direction only. Spend kept on the reviewed accounts is 98%, give or take 10 points, against 102% for the business’s other existing customers (same ledger report, new customers left out). A change under 14 points by the re-measure could be chance; one lost account (the retail park, at retender) explains most of the gap. Reactive sales rose 7%: better planned maintenance should lower them, so a fall would not be bad news. For scale only, a 5-point rise in spend kept would be worth £61,825 a year in gross profit and one more account of median size £31,500; neither is a forecast. The first year costs £17,275 in fee, set-up and 323 review hours.
Sales come from the sales ledger, service figures from the job system (planned visits and call-outs) and the complaints log and contacts from the Contacts log by role and date. The main measures are counts per account that the rhythm itself moves; with 24 accounts in the programme each is a small number, so read the counts, not just the percentages (two or more contacts: give or take 19 points). Spend kept compares the same 25 accounts over two 12-month periods (a lost account at what it bought before it left), give or take 10 points, so it shows direction only; reactive sales can fall because service improved. The comparison group (15 accounts) was chosen by alternating down the sales list below the accounts that must be reviewed, so it is fairer than choosing by size but still small, with no counterpart for the largest accounts. Only 12 renewals fall due before the re-measure. The target of 104% is the owner's assumption, used only as an illustration; gross profit is estimated from the margin for 16 accounts; review hours are built from the team's estimates of who attends and for how long. The re-measure nets spend kept against other existing customers from the same report, and shows whether the measures moved, not proof that the reviews moved them.
Fictional, continuing the example: Brynlowe Building Services, its customers and every figure here are invented, and customers’ people are named by role. The counts were tallied from the story, not by rerunning the workbook; seven of the agent’s log rows are on the Agent sheet of the worked-example workbook.
Brynlowe chose the guided route. At the final session on 20 October the account review agent was set up in an AI assistant the business already uses, with its IT’s agreement: connected read-only to the sales ledger, the job system, the complaints log and the account managers’ calendars, and writing only to the live copy of the workbook and one shared drafts document. The workbook as handed over was kept, read-only, as the baseline copy. The thresholds are on Inputs: a Draft pack for every review within 30 days (B30), Ready 2 working days before the meeting (B40), and a warning 30 days before a hold-by date with no review booked (B41). The run details are on the Agent sheet: it runs weekly (Monday 7:00), the managing director reads the note, and Gareth P approves anything that mentions a price. Each Monday it started the packs that had come inside the window, from the latest figures, and each working morning it looked for a review held the day before to draft the follow-up email. The account managers wrote the value to the customer and the evidence, marked each pack Ready, and sent every pack and email themselves. Nothing left the business, and no decision changed, without one of its own people approving it. At the check-in on Wednesday 25 November, 30 days after the agent’s first run, Geoff read the log with the managing director; his upkeep of the set-up ended there.
| Date | What the agent prepared or flagged | Approved or decided by | What happened |
|---|---|---|---|
| 26 Oct 2026 Draft | Housing association offices (6 sites) Started the Draft pack for the 18 November review: the year in numbers, what we delivered (both levels met) and the issues drafted from the ledger and the job system; the value and evidence left for the account manager. | Gareth P | Wrote the value section from the association's own count of tenants' heating complaints; Ready on 13 November, sent with the agenda on 16 November. |
| 2 Nov 2026 Flag | Leisure centres operator (3 centres) Hold-by date 30 November is inside the 30 days' warning and the next review isn't booked. Proposed 24 November from her diary. | Nadia R | Booked 24 November with the area manager and the operator's finance lead, before the hold-by date. |
| 2 Nov 2026 Flag | Academy trust (11 schools) The boiler replacement quotation, a next step due on 30 October, is still open. | Gareth P | Sent on 4 November and marked Done; it counts as late. |
| 5 Nov 2026 Draft | Care home group (8 homes) Follow-up email drafted from the three next steps agreed at the 4 November review; the operations director's meeting added to Contacts. | Tom H | Sent on 5 November. Two contacts engaged at the account, against one before. |
| 10 Nov 2026 Flag | Office landlord (5 buildings) The 12 November pack still in Draft on the day it should be Ready: value “TBC”, no evidence. | Gareth P | Wrote the value from the head of property's tenant survey and sent the pack that evening. |
| 23 Nov 2026 Flag | Food manufacturer (2 factories) A price (£2,400 a quarter) in the draft recommendations. No prices in the meeting. | Gareth P (director) | Price taken out. After the review, a quotation for a planned contract approved by Gareth P and sent on 2 December. |
| 18 Jan 2027 Draft | All accounts in the programme Early read: the six counts prepared from the live copy, the same way as the baseline, for checking. | Managing director | Checked and entered on the baseline copy's Value Case on 19 January. |
Ten of the eleven follow-up emails went within one working day; the car dealership’s went two days later, while Tom H was on leave. The leisure centres’ review was held before its hold-by date, which the calendar alone would have missed by three days. At the early read the managing director entered the six counts on the baseline copy: two or more contacts engaged 14 of 24 (from 11), next steps done by their due date 8 of 11, both service levels met 12 of 22 (from 11), and one of the 2 reactive-only accounts moved onto a planned contract (the builders’ merchant, from 4 January). Twenty-four accounts are too few to tell the rhythm from chance; the counts are read again at the re-measure. Finishing a Tier A pack from the agent’s draft took the account managers about 1.5 hours instead of 2.5 (their estimate), and reading the weekly note took the managing director about 10 minutes. The months after add nothing to the value case above.