Short answer
To know whether digital transformation or AI is paying off, take a baseline before each change, pick two or three business measures per initiative (time, cost, errors, cash or risk), measure them the same way afterwards, and report the difference in money and hours. Define the expected impact before you start, not after. Activity measures, such as users onboarded or bots deployed, show progress but not value.
Why is measuring value so hard?
Most organisations do not set the measure before they start. BCG's research published on 22 July 2026 found that nearly nine in ten CEOs say their companies see cost or revenue benefits from AI in targeted areas, and more than half cite linking AI initiatives to the P&L as a key barrier, yet only 14% have clearly defined the P&L impact for all AI initiatives, a gap of 42 percentage points (BCG, 22 July 2026). BCG adds that getting value requires discipline "from governance and workflows to accountability and value measurement".
Other reasons are familiar to any finance team: benefits are spread across many small tasks, time saved is not automatically money saved, and several changes happen at once, so it is hard to say which one moved the number.
What should you measure?
Choose measures that the board already understands. The table shows common ones and how to turn them into money.
| Measure | Example | How it becomes money |
|---|---|---|
| Cycle time | Days from purchase request to order | Faster supply, fewer rush costs |
| Staff time | Hours per month on invoice matching | Hours × loaded cost, if the time is redeployed |
| Error and rework rate | Invoices corrected after posting | Rework hours plus write-offs |
| Cash | Days sales outstanding | Working capital released |
| Customer outcome | First-response time; complaints | Retention and repeat revenue |
| Risk and control | Open audit findings; overrides; incidents | Expected loss avoided |
Pick two or three per initiative. More than that and nobody will read the report.
How do you set a baseline?
- Measure before you change anything. Use four to eight weeks of real data where possible, from systems rather than estimates.
- Write down how you measured. The same definition and source must be used afterwards, or the comparison is meaningless.
- Agree the expected result. For example, "approval cycle time falls from five days to two within three months of go-live". This is the P&L definition BCG found most companies lack.
- Name who will use the benefit. If staff time is saved, say what it will be used for. Otherwise it disappears.
Our guide on which processes to automate first recommends the same baseline step for every candidate.
What should you avoid when measuring value?
- Counting activity as value. Licences bought, users trained and workflows built are inputs. Report them, but do not call them benefits.
- Estimating instead of measuring. Vendor calculators and survey guesses are useful for a business case, not for showing results.
- Changing the definition halfway. If the way you count cycle time changes, restate the baseline too.
- Claiming every saving. If two initiatives touch the same process, split the benefit or report it once.
- Ignoring the cost of running the change. Support, review time and subscriptions continue after go-live and belong in the comparison.
How should you report progress to the board?
Use a one-page scorecard per quarter. For each initiative on the roadmap: the baseline, the target, the current value, the money or hours equivalent, and a one-line comment. Separate three kinds of result:
- Delivered: measured change against the baseline.
- Leading indicators: adoption, share of transactions going through the new process.
- Risk reduced: controls added, findings closed, incidents avoided.
Be honest about what has not worked. A board that sees one initiative stopped for lack of value will trust the others more.
Update, 29 July 2026: IBM's 2026 Cost of a Data Breach Report found that companies using AI and automation in security operations cut breach costs by almost $2 million on average (IBM, 29 July 2026). Risk reduction can be valued too, using expected loss avoided rather than cash saved.
How do you measure the value of AI specifically?
The same way, with two extra checks. First, measure quality, not just speed: an AI that drafts replies faster but needs heavy correction may save nothing. Track the share of outputs accepted without change. Second, count all the costs, including licences, usage charges, review time and the work of keeping data and prompts current.
Measure at the process level, not the tool level. "Hours saved by the AI tool" is hard to verify. "Time from enquiry to quote fell from three days to one" is not.
Update, 25 August 2026: McKinsey's 2026 State of AI survey found that 37% of respondents attribute at least some EBIT impact to AI, essentially unchanged from the year before, and that AI high performers are twice as likely as others to report defined processes to measure the impact of their AI initiatives (McKinsey, 25 August 2026).
Linkgrow's digital maturity audit sets baselines and measures alongside the roadmap, so results can be shown later. The free gap finder is a quick first step.
Frequently asked questions
What KPIs measure digital transformation success?
Business measures such as cycle time, staff hours, error rate, days sales outstanding, customer response time and open control findings, each compared with a baseline taken before the change.
How do you calculate ROI on automation?
Compare the measured benefit, such as hours redeployed times loaded cost, fewer errors and cash released, with the full cost, including licences, implementation, review time and maintenance.
Why do AI projects struggle to show value?
Often because the expected impact was never defined, no baseline was taken, or the AI was added to a process without redesigning it.
How often should we report transformation results?
Quarterly to the board, with a one-page scorecard, and monthly within the management team for active initiatives.
Sources
Where are the gaps in your operations?
Find your gaps