Commercial cleaning and planned building maintenance for offices, schools and surgeries across Shropshire and the Marches. 64 staff. Assessed by the owner with the operations director, with a staff survey.
| Dimension | Score | Level | Staff positive |
|---|---|---|---|
| Direction | 2.25 | Emerging | Not surveyed |
| People and skills | 2.00 | Emerging | 32% |
| Data and knowledge | 2.25 | Emerging | 47% |
| Systems and security | 2.25 | Emerging | Not surveyed |
| Processes | 2.50 | Established | Not surveyed |
| Rules and risk | 2.00 | Emerging | 18% |
| Culture | 2.25 | Emerging | 46% |
The weakest areas are people and skills and rules and risk, at 2.00. "Staff positive" is the share of survey answers that support readiness, where the survey covers that dimension.
Ranked by points short of 4 × weight, with open gates counted double. Owners and dates were agreed by Corvedale's owner.
| # | Gap | Action | Owner | By when |
|---|---|---|---|---|
| 1 | S2 Are AI tools used through business accounts rather than personal ones? | Move AI use onto business accounts and stop personal accounts for work | Operations director | November 2026 |
| 2 | G1 Are there written rules on what may and may not be put into AI tools? | Write and share a one-page rule on what may and may not go into AI tools | Owner | November 2026 |
| 3 | P2 Can the people using AI write good instructions and check the answers? | Run short practical training for the pilot group on instructions and checking | Operations director | January 2027 |
| 4 | A2 How reliable is that information? | Clean up duplicate customer records and name an owner | Office manager | December 2026 |
| 5 | D2 Are AI efforts tied to specific business goals? | Set two goals with measures: hours on job reports and time to answer customer queries | Owner | November 2026 |
This assessment rests on 28 scored questions: 11 backed by a document or setting seen, 4 observed, 4 from the staff survey and 9 on the owner's view alone. 17 of 22 staff responded to the survey. The thresholds are starting assumptions, not industry statistics.
The evidence is thinnest in rules and risk (3 of 4), direction (2 of 4) and processes (2 of 4), where scores rest on the owner's view alone. These 9 owner's-view scores should be firmed up as the actions are done. Contracts with schools and surgeries have not yet been checked for limits on data use, so the client-duties score (G3) should be revisited once they are.
Once the two open gates are closed, Corvedale is ready to value specific tasks with the AI Opportunity Map. The time-consuming job reports and quotations are the obvious places to start.
Fictional and illustrative. This continues the worked example from the survey in late September 2026 to January 2027. Every name and figure is invented.
Corvedale set up the readiness agent in its own accounts at the first working session, before the survey went out: its AI assistant, its SharePoint and the form tool. The rules are on Inputs, rows 26 to 31. Dawn Pritchard, the office manager, sent the survey and its one reminder and held the raw answers; the form recorded no names. It stayed open 7 days, with the reminder after 4. Gareth Lloyd, the operations director, confirmed every piece of evidence the agent found, and anything older than 12 months was flagged. The owner and Gareth chose every score. The signed-off assessment is unchanged; the agent keeps a working copy for the April 2027 repeat.
During the fieldwork the agent drafted 2 messages, and Dawn read and sent both. Nothing went to staff that Dawn hadn't read and sent herself, and the agent never saw a single answer, only the totals. After handover it checked the evidence folder and the action dates each Monday and wrote a short note to Richard.
| Date | What the agent found | Decided by | What they decided |
|---|---|---|---|
| 29 Sep 2026 | Survey invitation drafted: the eight statements, no names asked Survey guide; Inputs B26 to B28 | Dawn Pritchard | Dawn read it, added a line about Friday's supervisors' meeting and sent it to the 22 staff on 29 Sep. |
| 3 Oct 2026 | Responses after 4 days: 9 of 22 (41%) Forms response count | Dawn Pritchard | One reminder drafted for everyone invited (the form records no names, so nobody is singled out). Dawn sent it on 3 Oct. |
| 6 Oct 2026 | Survey closed: 17 of 22 responded (77%) Forms response count; 17 is above the minimum of 5 on Inputs B15 | Dawn Pritchard | Statement counts entered on the Survey sheet. Raw answers stay in Dawn's Forms account until the assessment is signed off, then deleted. |
| 6 Oct 2026 | 8 documents and settings listed against 7 questions (D1, D3, P4, A3, S1, S3, G2) Shared drive: board minutes, IT provider's report, system settings, record of processing | Gareth Lloyd | Gareth confirmed 7 and dropped 1 (a draft IT policy never adopted). The owner and Gareth chose the scores; the agent chose none. |
| 7 Oct 2026 | Evidence older than 12 months: the operations manual for R1 is dated March 2025 Shared drive: Operations > Manual; Inputs B30 = 12 | Gareth Lloyd | Gareth walked through job booking with the office team: the manual still matches. Kept as Seen, with the check noted in the evidence. |
| 16 Nov 2026 | One-page AI rule saved (G1 action, due November 2026) Shared drive: Policies > AI use rule, 12 Nov 2026; briefing note, 13 Nov | Richard Bevan | Richard confirmed it was read out at the 13 Nov staff briefing. Logged for the repeat; the scores stay as assessed until then. |
| 7 Dec 2026 | IT provider's report shows no personal AI accounts on work laptops (S2 action) IT provider's monthly report, 1 Dec 2026 | Richard Bevan | Both open gates now have evidence logged. Richard kept the April 2027 repeat: the briefing and training need time to show in the survey. |
| 4 Jan 2027 | A2 action past its date (December 2026) with no evidence logged Assessment sheet, By when; evidence folder | Richard Bevan | Dawn has until the end of February: the duplicate customer records are being merged one contract type at a time. |
By the end of January, both open gates had evidence logged: the one-page rule on what may go into AI tools, read out at a staff briefing, and the IT provider's report showing no personal AI accounts on work laptops. Two goals with measures were minuted in November and the pilot group's training was held on 20 January, so 4 of the top 5 actions had evidence. The duplicate customer records ran late and have until the end of February. None of this changes a score yet: evidence logged is not a gate closed until the repeat scores it. Richard kept the April repeat rather than bringing it forward, because the briefing and the training need time to show in what staff say. Dawn spent about 10 minutes reading and sending each draft, and Gareth about 15 minutes a week confirming evidence (estimates). The April repeat starts with the evidence already gathered and dated.