Salesforce Lightning · Agentforce · Copilot · any AI tool
The tool went live. Did your people?
Salesforce changes three times a year, and AI is now arriving inside it and inside everything else. Most businesses buy the tool and skip the part where people change how they work. That part is ours.
Why it stalls
Everyone is buying AI. Few change how the work gets done.
Four dated sources. The last two columns are our reading, not a finding.
| What the data says | When | What it means for you (our reading) | What we do about it |
|---|---|---|---|
| 88% of organisations use AI in at least one function. About a third have begun to scale it. Around 6% report real profit impact, and they are nearly three times as likely to have redesigned workflows.Source type: consultancy survey of 1,993 organisations (McKinsey) | Fieldwork June to July 2025, published 5 November 2025 | Owning the tool is not the gap. Changing the work is. | We rebuild training around the real tasks in each role, not the menu of features. |
| 35% of UK businesses with 10 or more employees used AI in June 2026, up from about 12% in 2023. Only 11% said over half their workforce had AI-related training.Source type: official statistics (ONS business survey) | Data to June 2026, published 20 July 2026 | Use is running ahead of training. | Training by role, with clear rules on what people should and should not put into AI tools. |
| Gartner expects over 40% of agentic AI projects to be cancelled by the end of 2027, citing cost, unclear value and weak risk controls. In July 2026 KeyBanc analysts said Agentforce customers’ data is not ready.Source type: analyst forecast (Gartner, as reported by New Electronics) and analyst commentary (CIO) | Gartner June 2025; KeyBanc via CIO, 17 July 2026 | An agent is only as good as the data and process behind it. Check both before go-live. | We run real cases through it before go-live and hand the gaps to the people who own the data and the configuration. |
| Salesforce ships three major releases a year, and its own admin guidance says to teach users the new features each time.Source type: vendor guidance (Salesforce admin blog) | Every release: Spring, Summer, Winter | Training is never finished at go-live. | A short briefing per role after each release, then a usage check. |
Sources: McKinsey; ONS; New Electronics on Gartner; CIO on KeyBanc; Salesforce admin blog.
What we hear
Four sentences we hear before we are hired
“We bought it. Nobody uses it properly.”
The licences are live and the old spreadsheets are still open.
What we do Check usage role by role, find where people stall, and fix the job aid, not the person.
“Training was a vendor webinar.”
Generic courses teach the product. Your people need your lead stages, your hand-offs and your rules.
What we do Train on your own org with your own records, built from your design documents and procedures.
“People are already using AI. We do not know how.”
Staff are often ahead of the policy.
What we do Agree the approved uses for each role, then train to them.
“The agent worked in the demo.”
Live data and live cases are messier than a demo.
What we do Test with real cases before go-live and hand the gaps to their owners.
How we fix it
Four steps, in the order that matters
- Pick five tasksThe real tasks in each role, not the feature list.
- Train on your orgYour records, your sandbox, your rules.
- Stay on the floorSupport at go-live and a short list of the real questions.
- Check usageAfter go-live and after every release, role by role.
What we do not do. We do not build the AI or configure Salesforce. We make sure the people using it can, and we pass configuration gaps to whoever owns them.
Next step
Send three lines. We will tell you where it will stall.
Tell us what is going live, when, and who will use it. We reply with the three places we would look first.