ChatGPT
Most ChatGPT rollouts start well: people already know the tool, so logins arrive quickly. Then usage splits. A few people build clever GPTs, most use it as a smarter search box, and nobody can say which workflows changed. The problem isn't the model. ChatGPT was handed out, not designed in.
15 minutes. No sales pitch.
We don't run prompt training. We work with individual teams to rebuild three or four real workflows around ChatGPT: the context it needs, the approved sources it should use, where it drafts and where a person decides. Then we measure those workflows by cycle time and quality, not by weekly active users.
| Team | Task | What ChatGPT does | Feature |
|---|---|---|---|
| Finance | Variance commentary | Analyses an uploaded export and drafts commentary for review | Projects, data analysis |
| Finance | Supplier or market checks | Produces a sourced research summary | Deep research |
| Operations | Standard responses and procedures | A GPT holds the approved procedures and drafts consistent answers | Custom GPTs |
| Sales | Account preparation | Pulls notes and documents from connected tools into a brief | Apps |
| HR | Policy questions | A workspace GPT answers from the approved handbook only | Custom GPTs |
| Legal / professional services | First-pass review against a checklist | Lists issues against your standard positions for a lawyer | Projects, custom GPTs |
On Enterprise, workspace analytics shows adoption and engagement trends by group, with CSV exports and impact surveys. It refreshes within 24 hours and doesn't expose message text. The Compliance Platform provides logs for eDiscovery and security tools. Neither shows whether a draft was right or a process got faster.
We baseline each target workflow first (time, rework and a quality check), then measure again after the change. Analytics tell you who's using it. Our measurement tells you what it changed.
Illustrative scenario
This is an illustrative scenario, not a client engagement. A 600-person manufacturer has ChatGPT Business across its commercial and HR teams. Dozens of overlapping GPTs exist, and several staff still use personal accounts. We map which GPTs people actually use and retire the duplicates. Then we rebuild two workflows, quote-request responses and HR policy answers, each around one approved GPT with a named owner, and measure response time and rework against a baseline.
Twisthand Intelligence is a Nottingham-based consultancy that helps UK organisations turn ChatGPT Business and Enterprise into redesigned, measured workflows, team by team, delivered hands-on.
OpenAI says that by default it doesn't use data from ChatGPT Enterprise, Business or Edu to train or improve its models.
OpenAI offers UK data residency for storage at rest on eligible Enterprise workspaces, and it's rolling out on Business. In-region processing currently covers Europe, the US and the UAE, not the UK. Check what your workspace has.
Enterprise adds controls such as role-based access, GPT sharing controls, the Compliance Platform, workspace analytics and data residency options.
Give each approved GPT a named owner, retire duplicates, and on Enterprise use permissions to control who can create and share them.
Copilot is built into Microsoft 365 and grounded in it. ChatGPT is a separate workspace that connects to Microsoft 365 and other tools. Some organisations use both for different jobs.
Both are capable, and both exclude business data from training by default. Choose on the workflow, the tools you need to connect, data-location requirements and admin controls, then prove it on one workflow.
Fifteen minutes, no slides, no pitch. Tell us where ChatGPT has stalled and we'll tell you honestly where it could pay back.
Book a callOr see engagements & pricing.