StarLight · Deep Field
AI Adoption
User: Grace Oyelaran
Account Manager · Sales · 19 days worked
Sep 3 – Oct 2, 2026
By tool
Answers “Claude or ChatGPT?” for this person
| Tool | Prompts | Conversations | Tasks delegated | Share of prompts |
|---|---|---|---|---|
| Claude (web) | 46 | 18 | 8 | 40% |
| ChatGPT | 31 | 14 | 5 | 27% |
| Claude Code | 21 | 9 | 7 | 18% |
| Our platform | 12 | 6 | 3 | 10% |
| Gemini (web) | 6 | 4 | 0 | 5% |
Trend
Each day worked in the range
What the conversations were for
51 conversations. Automate means the AI did the work. Assist means the person did it with help.
Automate: 23 (45%), which are the tasks delegated · Assist: 24 (47%)
- Dark Agents 3
- Human-in-loop 20
- Ask 12
- Advise 5
- Learn 4
- Check 3
- Not judged yet 4
Tasks delegated
23
AboveCompany avg 11
+9 vs Aug 4 – Sep 2
Prompts per day
6.1
AboveCompany avg 4.8
+1.3 vs previous period
Prompts per task
3.4
BelowCompany avg 5.0
−0.8 vs previous period
Tasks per day
1.2
AboveCompany avg 0.6
+0.5 vs previous period
AI maturity ladder
Where this person sits, based on how they use AI
- 4
Autonomous
Hands off tasks that run without them
- 3
DelegatingGrace
Hands tasks to AI and steers it
- 2
Assisted work
Gets help on their own tasks
- 1
Asking
Questions and learning only
- 0
None
No AI use in the range
Moves up to Autonomous when at least a quarter of delegated tasks run without them steering. That's 13% now (3 of 23).
Briefing quality by tool
Prompts each delegated task took, and how often they re-asked the same thing
| Tool | Prompts per task | Re-ask rate |
|---|---|---|
| Claude Code | 2.0 | 3% |
| Claude (web) | 3.0 | 6% |
| Our platform | 4.7 | 8% |
| ChatGPT | 5.2 | 16% |
| Company avg | 5.0 | 9% |
Shows where they brief AI well and where they go back and forth. Coaching material, not a score.
Value of AI time by tool
Conversations by value class
- Claude (web)18
- ChatGPT14
- Claude Code9
- Our platform6
- Gemini (web)4
- Producer
- Contributor
- Maintainer
- Distractor
Risk by tool
Waste, data leaks and unapproved tools. Each count links to the AI Prompts page.
| Tool | Waste | Data leaks | Status |
|---|---|---|---|
| Claude (web) | 2 (4%) | 0 | Approved |
| ChatGPT | 3 (10%) | 1 | Approved |
| Claude Code | 0 | 0 | Approved |
| Our platform | 1 (8%) | 0 | Approved |
| Gemini (web) | 2 (33%) | 0 | Unapproved |
When they use AI
Prompts by hour and weekday, in company time
AI-active days
Days with any AI use, out of days worked
16 / 19
days worked used AI
7 days
longest streak
- Used AI
- Worked, no AI
- Not worked
Recent delegated tasks
Only for viewers who can already read this person's prompts. Everyone else sees counts.
Draft the Q4 renewal email sequence for the Apex account from these notes
Claude (web)4 prompts
Build a pivot of September pipeline by stage and flag deals idle > 14 days
Our platform6 promptsPipeline review → Weekly roll-up
Refactor the quote calculator script so discounts stack correctly
Claude Code2 prompts
Summarise the three competitor pricing pages into a comparison table
ChatGPT7 prompts
Write follow-ups for everyone who attended Tuesday’s webinar
Our platform3 promptsWebinar series → Follow-ups
Turn this meeting transcript into action items with owners
Claude (web)2 prompts
New tools this period
First used in the range, with nothing in the previous period
Claude Code· first used Sep 22
21 promptsGemini (web)· first used Sep 29
Unapproved
Waiting on AI
Average time from sending a prompt to typing again
| Tool | Average wait |
|---|---|
| Claude (web) | 41s |
| ChatGPT | 55s |
| Claude Code | 18s |
| Our platform | Not measured |
| Gemini (web) | 38s |
Captured only on desktop, so our platform has no figure. Breaks longer than 60 minutes count as stepping away and are left out.
See your AI value in 24 hours. Get the full strategy in 15 days.
Free. No obligation. No integrations.