Agentic ROI Calculator · Deep Hospitality

Choose a processStart from a hotel example, then make it yours.

CurrencyExample figures are illustrative. Replace them with your own.

The work todayWho does it, and how long it takes.

People involved

Hours each, per week

Loaded cost per hour

GBP

Share the agent can take (typical 35 to 65%)

Working weeks a year

What the agent changesCount only what would really land on the P&L.

Money that leaves the business

External spend on this process, per year (agencies, outsourcing)

GBP

Share the agent could avoid

Cost of delay

Weekly cost while the work waits (overtime, cover, lost sales)

GBP

Weeks of waiting removed, per year

Revenue you miss today

Opportunities a month (enquiries, RFPs, arrivals)

Missed or slow today

Share the agent recovers

Conversion rate

Average value

GBP

Flow-through to profit

Errors, waste and penalties

Avoidable cost per year

GBP

Count time saved as cash
Only tick this if you will genuinely buy fewer hours (less overtime, temps or agency cover). Otherwise it is capacity: time back for guests and teams, never a headcount saving.

What it costsIncluding the people who check its work.

One-off

Build and set-up

GBP

Knowledge library (folders, rules)

GBP

Training and change

GBP

Running

Licensed users

Licence per user, per month

GBP

Usage and platform, per month

GBP

Human oversight, hours a week

Support, % of build a year

Oversight is the cost most business cases leave out: someone approving, checking and correcting. It is costed at the loaded hourly rate above.

Readiness and riskBe honest. It changes the answer.

Is the work ready? Four questions

IntentWe know the outcome we want, and how we will measure it. ProcessThe steps are mapped, exceptions included. DataThe knowledge it needs is written down, current and in one place. OwnerA named person is accountable for the result.

Adoption ramp

Year 1

Year 2

Year 3

Confidence

Optimistic · 0%Cautious · 40%

Cost contingency

Discount rate

How the calculator works

Cash and capacity, kept apart

Most AI business cases turn every minute saved into money. Owners see through that. We count four kinds of cash that really move: external spend avoided, the cost of work waiting, revenue you currently miss, and avoidable errors or waste. Time saved is shown as capacity, and only counts as cash if you will genuinely buy fewer hours.

Every cost, including people

Licences and usage are the visible costs. Build, the knowledge library, training and ongoing support follow. Then the one most cases leave out: the hours people spend approving and checking the agent's work. A good agent keeps people in every decision that matters, and that time is real.

Ramp, risk and readiness

Benefits ramp up over three years rather than arriving on day one. A confidence haircut reduces benefits, and contingency increases costs, in the spirit of Forrester's Total Economic Impact method. Each readiness question you cannot yet answer adds a further 10% haircut, because an unclear process automated is still an unclear process.

The maths

Cash benefit a year = external spend × share avoided + weekly delay cost × weeks removed + missed opportunities recovered × conversion × value × flow-through + errors avoided
Capacity = people × hours × weeks × share the agent takes
Year benefit = full benefit × adoption × (1 − haircut)

The outputs

Net value: three years of benefits minus all costs
NPV: the same, discounted each year
ROI: (PV of benefits − PV of costs) ÷ PV of costs
Payback: the month cumulative cash turns positive

Use it well

Pick one process and one P&L line it lands on
Set a baseline before you start, and a named owner
Review at 90 days: scale, hold or stop
Treat the answer as a range, not a promise

What we drew on

Forrester, Total Economic Impact of Microsoft 365 Copilot (March 2025): three-year, risk-adjusted benefits and costs, discounted at 10%.
Salesforce, Agentforce ROI calculator: benefits ramp over three years, and implementation costs are often excluded. We include them.
Kanerika, agentic AI ROI calculator: typical efficiency of 50 to 65% for high-volume tasks and 35 to 45% for exception-heavy work.
Hotel Online, Put AI on the P&L: hours saved only count when they cut purchased hours, win sales or avoid a cost; count oversight hours.
Gartner (June 2025): over 40% of agentic AI projects are predicted to be cancelled by the end of 2027, citing cost and unclear value.
BCG, AI disruption in hospitality (2026): AI leaders see larger revenue and cost gains; treat projections as ranges.

Illustrative only. This is not financial advice, and your results will differ. Nothing you enter leaves your browser.