To measure AI ROI in an agency, track three layers against a baseline taken before rollout: hours saved on repeating workstreams, the cost and capacity of pitching, and revenue or margin from the time you redeploy. Then subtract the full cost of tools and training. Hours alone are not ROI; they become return only when they are billed, used to win business or taken out of cost.
Key takeaways
- Nearly half of agencies do not measure AI ROI yet, and almost as many measure only time saved.
- Measure in three layers, hours, pitches and revenue, each against a number recorded before the rollout.
- SOMIN's agency case studies show the scale of the first layer: monthly reporting from 9 hours to 1 and brand analysis from 10 hours to 1.
- Saved hours become return only when they move to billable work, more pitches or lower cost, so track where they go.
- Report AI ROI the way clients judge agencies: in business outcomes, not activity counts.
Why do most agencies struggle to measure AI ROI?
Because they skip the baseline, measure activity instead of outcomes and bill in a way that hides the gain. Without a before number, every after number is an anecdote.
EMARKETER reported in January 2026 on AI Digital's agency AI adoption benchmark: 46.4% of agencies did not measure AI ROI yet, and another 46.4% measured it only through time savings and efficiency. Outside agencies the picture is similar. The State of AI 2026 survey from McKinsey, covering 1,719 participants, found 37% attributed any EBIT impact to AI, and only 6% qualified as high performers attributing 5% or more of EBIT to it.
Billing makes it harder. A Forrester analyst told Digiday in May 2025 that "AI has not as of yet transformed the remuneration model". If you bill by the hour and AI halves the hours, revenue falls unless the time goes somewhere else.
The pressure to prove impact is visible in SOMIN's data too. In our audience-research workspace, a study of two years of organic posts from audience-research brands (191 posts, April 2026) found 19 posts on one theme, "Transforming Marketing ROI Measurement": the need for simpler, more effective ways to show what campaigns return.
How many hours does AI save an agency?
Time one repeating workstream per cycle, before and after, in the same unit: per pitch, per client or per month. In SOMIN's published agency case studies, research and reporting show the largest drops.
Click2View Asia runs competitor monitoring and reporting on SOMIN. Monthly client reporting fell from 9 hours to 1, an 89% reduction; a comprehensive competitive analysis fell from 18 hours to 3; and content strategy groundwork from 14 hours to 2, according to the Click2View Asia case study. Headcount per workstream went from two people to one.
Blak Labs, a 50-person Singapore agency, cut brand analysis per pitch from 10 hours to 1, content planning from 10 hours to 3 and campaign planning from 5 hours to 2, as the Blak Labs case study reports.
Does AI change the economics of pitching?
Yes, and pitching is often where ROI is easiest to prove, because pitch work is non-billable by definition. Track cost per pitch, people per pitch, turnaround and how many pitches you can afford to run.
- KPI Media moved pitch preparation from a one-week sprint to two days, from three people to one, and from $4,000 to $1,000 per pitch, a 75% cut (KPI Media case study).
- NEO360 consolidated proposal prep from a media buyer, an analyst and an account manager to the account manager alone, cutting turnaround from 4 hours to 2.5 and cost per proposal from $1,000 to $700, a 30% reduction (NEO360 case study).
- Havas Media's Middle East performance practice uses SOMIN to give new clients a factual category read from day one, rather than after hours of manual research (Havas Media case study). Its measure is speed to strategy, not a cost figure, and that is a valid KPI too.
When does AI show up in revenue and margin?
When saved hours are turned into won business, upsell or lower cost on a workstream. This layer is what makes the case to a managing partner or a CFO.
KPI Media closed at least one new client off its faster pitch process, adding $5,000 a month in retainer revenue. Blak Labs reduced the effort on client reporting from 4 FTEs to 1.75, with monthly salary cost on that workstream falling from $10,000 to $6,000. Click2View Asia projects a 100–400% new-business uplift and a 100–200% upsell to existing clients from the reclaimed senior time; those are the agency's projections, not results yet, and should be reported as such.
Notice that the revenue gains follow redeployment. NEO360 did not cut its media buyer or analyst; it moved them back onto billable client work, which is why the saved hours turned into output.
Which KPIs should an agency track for AI ROI?
Eight KPIs cover the three layers. Record each one before rollout, then monthly.
| Layer | KPI | How to measure it | Published example |
|---|---|---|---|
| Hours | Hours per cycle | Hours on one workstream per month or per pitch, before and after | Click2View: monthly reporting 9h → 1h |
| Hours | People per deliverable | Named contributors per pitch, proposal or report | KPI Media and NEO360: 3 → 1 |
| Pitches | Cost per pitch | Hours × loaded hourly cost, plus a share of tool cost | KPI Media: $4,000 → $1,000 |
| Pitches | Turnaround | Elapsed time from brief to finished pitch or proposal | KPI Media: 1 week → 2 days |
| Pitches | Time to first strategy | Days from signed client to an agreed media or content strategy | Havas Media: category read from day one |
| Revenue | Workstream cost | Monthly salary cost allocated to the workstream | Blak Labs: $10,000 → $6,000 |
| Revenue | New revenue attributable | Retainers or upsell where AI-built work was part of the win | KPI Media: +$5,000 a month |
| Revenue | Redeployment rate | Share of saved hours moved to billable or new-business work | NEO360: specialists back on client work |
How do you turn the numbers into an ROI figure?
Use a simple formula: ROI equals the value of redeployed billable hours, plus new margin, plus cost removed, minus the total cost of AI, divided by the total cost of AI. Count hours only when they were billed or used on new business.
On KPI Media's published figures, for example, each pitch costs $3,000 less to prepare and one retainer won through the new process adds $5,000 a month. Those two lines, set against the platform and training cost, are the whole ROI conversation, and they fit on one slide.
Frame the result in the client's language as well. SOMIN's moments-of-truth model, which generates the micro-moments a persona is likely to act on, produces a client-side thought: "I need to evaluate my agency based on real results, not just ROAS or CPA." It is model-generated, not a quote, and it pairs with a SOMIN insight on opportunity cost: teams that chase vanity metrics misallocate resources away from real profit. An agency that reports its own AI ROI in outcomes earns the right to report campaigns the same way. More on building that offer is on our SOMIN for agencies page.
FAQ
How do you calculate the ROI of AI in a marketing agency?
Take the value of redeployed billable hours, new margin and cost removed, subtract the total cost of AI tools and training, and divide by that cost. Use a baseline recorded before rollout for every input.
What KPIs show whether AI is working in an agency?
Hours per cycle, people per deliverable, cost per pitch, turnaround, time to first strategy, workstream cost, new revenue attributable and the share of saved hours redeployed to billable work.
Is time saved enough to prove AI ROI?
No. Time saved is an input; it becomes return only when the hours are billed, used to win new business or removed from cost. Track where the hours go as carefully as how many you save.
How long before an agency sees ROI from AI?
Hours and cost per pitch can move within the first few cycles of a workflow. Revenue effects take longer because they depend on pitch outcomes and retainers, so track all three layers from the start.
Before you change anything else, record a baseline for one workstream and one pitch: hours, people, elapsed time and cost. Re-measure after 30 and 90 days, write down where every saved hour went, and take the eight-row table to your next partners' meeting.
