Prepared for Operations leadership · November 2026
Your CEO calls AI "the biggest change since the barcode"[A]. What you report today is log-ins.
Your annual report sets the ambition and names two AI use cases[A]. You told us your dashboard shows licences and monthly active users. Nothing yet shows whether your teams can use AI well, and that is the number your CFO will ask for.
- Company
- Quillhaven Logistics (fictional)
- Industry
- Logistics and supply chain
- Written from
- A fictional company
What this report is built from
- You told us7 from your call
- We found5 public facts
- Published research3 studies
- Our assumption6 estimates
The leading indicator
- Input
Seats
Do your people have AI?
640[B]
office staff licensed for one approved assistant
We found
- Activity
Usage
Are they using it?
71%
monthly active, reported to the board each quarter
You told us
- Leading indicator
Capability
Can they use it well?
Not measured
no view of capability by team
You told us
This report focuses here
- Lagging indicator
Outcomes
Is it paying off?
2[A]
use cases named in the annual report, no result published
We found
- 72%
Because your CFO asks what the programme returns
of senior leaders at large US companies track structured, business-linked ROI metrics for AI.[1]
- 19%
Because you report monthly active users
of US employees use AI at work a few times a week or more.[2]
- 14%
Because customer service is where you start
more issues resolved per hour by customer support agents given an AI assistant, and 34% more for novices.[3]
Where AI value sits in road freight
- Cost
Customer service
Customer answers from the order record
Answer track-and-trace and booking questions from the order record.
PromptingModel core
- Cost
Customer service
Delay updates
Draft the customer update when a delivery slips, from the driver's scan and depot note.
PromptingWorkplace
- Revenue
Customer service
Rebooking offers
Offer a rebooked slot or a faster lane when a customer asks about a delay.
PromptingGovernance
- Risk
Customer service
Damage claim intake
Check a damage claim against the delivery proof before it reaches the claims desk.
AgentsGovernance
Capability by team
Comparable roles using AI often
Customer service27%
US white-collar employees[2]
IT and data50%
US employees in technology[2]
Regional managers33%
US leaders (managers of managers)[2]
Depot operations9%
US production and front-line workers[2]
The teams are invented for the sample; each figure is real published research. In your report the baseline measures capability itself.
Biggest gaps first
- 1
Model core in Customer service
Answers must match the order record, even when the assistant sounds sure.
- 2
Governance in Customer service
Rebooking offers carry customer data and pricing rules.
- 3
Agents in Depot operations
Claim checks hand a case from the depot to an agent.
What you told us, and what we found
What you told us
“the board asks what the programme returns and we only have usage”
- Customer service, 190 people, is where you want to start.
- A shift lead checks every answer the assistant drafts for a customer.
- The programme's budget is decided in the 2027 budget round.
What public sources say
- A
Stated aim:Digital operations is one of three investment priorities for 2026 to 2028.
Annual report 2025
- C
11 of 38 open office roles ask for experience with AI tools.
Careers page, November 2026
- D
AI training appears once in public reporting, as an intention for 2027.
Sustainability report 2025
These sources are invented for the sample. In your report each signal links to the public page it came from.
The value to work towards
This is where we would work towards with you: an estimate built from published research and the facts we found, not a measurement.
$85k to $188k
Gross annual value at stake, before programme costs
190 people × 27% = 51 people × $1,672 to $3,678 a year
Revenue$28k to $63k
Cost$57k to $125k
RiskNot priced
- People
- 190 people in the first group, as you told us.
- Who applies it
- 27% of US white-collar employees use AI at work a few times a week or more[2]: 51 people.
- Value per person
- $1,672 to $3,678 a year: 1 to 2.2 hours a week saved[4] at $36.34 an hour[5].
Three steps before the 2027 budget round
BaselineWeeks 1 to 2
- What you do
- Agree a sample of order questions, the review rule and the pass mark with the customer service lead.
- What to measure
- Answers the shift lead accepts unchanged, and the time each takes including checks.
BuildWeeks 3 to 8
- What you do
- The team practises daily on its own order questions, delay updates and rebooking offers.
- What to measure
- Capability by area for customer service, week by week, next to accepted answers.
ConnectWeeks 9 to 12
- What you do
- Read the same sample again and take both numbers to finance before the budget round.
- What to measure
- Accepted answers and time per answer, set beside the 71% monthly active use.
Week 12If the pass mark is met, run the baseline with depot operations; if not, review the sample with the shift leads first.
Next step
Agree the baseline sample with the customer service lead, then start week 1.
See it on two of your teams. A call with Alex van Gennep to scope a pilot: which teams, what you would measure, and what you would take to your CFO at the end.
kju.ai/ai-roi
Method and sources
- Capability is what people can do with AI in their roles, across five areas: model core, prompting, agents, governance and workplace. kju.ai has not measured your teams. The team figures are the closest published research, and the baseline measures capability itself.
- The value range is gross: the people who apply a use case, times a low and a high annual value per person, every input from published research. It leaves out programme costs and puts no dollars on risk. Every input is listed below.
- Capability is a candidate leading indicator: an early sign that the conditions for a return may be in place, not proof or a prediction of one.
- Model core
- knowing what a model can and cannot do
- Prompting
- asking for the output the work needs
- Agents
- handing a task to an agent safely
- Governance
- data, approvals and review rules
- Workplace
- AI in everyday tools, files and data
Sources
- 1Wharton Human-AI Research and GBK Collective, Accountable Acceleration: Gen AI Fast-Tracks Into the Enterprise, October 2025. About 800 senior decision makers at US companies with 1,000+ employees.
- 2Gallup, AI Use at Work Has Nearly Doubled in Two Years, June 2025. Quarterly survey of US employees.
- 3Brynjolfsson, Li and Raymond, Generative AI at Work, The Quarterly Journal of Economics 140(2), 2025 (NBER Working Paper 31161). 5,179 customer support agents.
- 4Federal Reserve Bank of St. Louis, The Impact of Generative AI on Work Productivity (Bick, Blandin and Deming), February 2025. Nationally representative survey of US workers, November 2024.
- 5US Bureau of Labor Statistics, Employer Costs for Employee Compensation, June 2026 (released 9 September 2026), private industry workers, total compensation per hour worked.
A to D: company sources, invented for this sample.
The value range, input by input
- People
- 190 people in the first group, as you told us.
- Replace with: The count your baseline confirms.
- Who applies it
- 27% are assumed to apply a use case: 51 people. Estimate from published research: 27% of US white-collar employees use AI at work a few times a week or more[2].
- Replace with: The share your baseline finds applying it.
- Value per person
- $1,672 to $3,678 a year for each of them: 1 to 2.2 hours a week saved[4], at $36.34 an hour for sales and office occupations[5], over 46 working weeks (our assumption).
- Replace with: Your loaded hourly cost, and the time saved after review that your baseline records.
- By lens
- The people applying are spread evenly over the 3 use cases that carry dollars: 17 people on revenue, 34 people on cost.
- Replace with: Where your baseline finds the time saved.
190 people × 27% = 51 people × $1,672 to $3,678 a year = $85,272 to $187,578
Risk use cases carry no dollars: what one is worth depends on the size of the exposure it reduces. This is gross capacity, not cash and not a business case. A business case subtracts the programme's cost and says how the time released is used.
Something here not right?
In your report, tell us what does not match and a person at kju.ai corrects it and sends a new version.
About kju.ai
kju.ai closes the gap between what AI can do and what your teams can do. People practise AI for their own role, one day at a time. What they learn rolls up into a view of capability by team and by area, which we call the AI ROI Leading Indicator. This report applies it to Quillhaven Logistics.
kju.ai · AI ROI Leading Indicator · Quillhaven Logistics (fictional company)