How to Measure ROI on the AI Tools You Already Pay For

Most teams know which AI tools are in the budget. Fewer know which ones are actually earning their place. A four-step process (baseline, track, calculate, decide) gives you a clear answer for every tool in your stack within 90 days.

Here is what this post covers:

Setting a baseline before you start measuring
Tracking the right inputs during the evaluation window
Calculating ROI in a format that informs a keep-or-cut decision

Cross-referenced against IBM Think Insights AI ROI guidance (ibm.com/think/insights/ai-roi) and SMB AI ROI frameworks on 2026-07-02.

Prerequisites

A list of the AI tools your team currently pays for (subscriptions, per-seat plans, or usage-based billing)
An estimate of the hourly cost of the employee time each tool is intended to save
About two hours to set up tracking, then 30 minutes every two weeks to update it

Step 1: List Every AI Tool With Its Monthly Cost

Open your finance system, card statements, or whatever your company uses to track SaaS spend. Pull every AI tool line item: writing assistants, meeting summarizers, code completion tools, CRM AI features, support ticket triage tools, and anything else with AI in the product description.

For each tool, record:

Tool name
Monthly cost (or annual cost divided by 12)
The team or individuals who use it
The task or workflow it is supposed to accelerate

This list is the starting point. You cannot measure ROI without knowing what you are spending.

Step 2: Set a Baseline for Each Tool

A baseline is the before-state: how long did the task take, or how often did it happen, before the AI tool was in use? Without it, you have no reference point and no way to calculate what changed.

For each tool, document:

The task it handles (specific, not general: not "writing" but "writing weekly status reports")
Time spent per task before the tool (ask the users who do this work)
Frequency (how many times per week or month)

If the tool is already in use and you do not have a pre-tool baseline, use the user's estimate of how long the task took when they were not using AI. Estimates are less precise than measured data, but they are good enough to produce a directionally accurate ROI calculation.

Step 3: Track Time Saved Over 30 Days

Ask each tool's users to log actual time saved on a simple shared document or spreadsheet for 30 days. Each row:

Date
Task completed with the tool's help
Estimated minutes saved vs. doing the same task without the tool

This does not need to be elaborate. A running note in Notion or a shared Google Sheet with five columns is enough. The goal is a real-world data set, not a perfect accounting.

If users resist the logging step, narrow the scope: instead of tracking all usage, track one representative task per person. A writing assistant used for proposal drafts, measured for one proposal per week, produces a clean data point that scales.

Step 4: Calculate Monthly Value

After 30 days of tracking, calculate the value the tool generated:

Total minutes saved per month across all users
Divide by 60 to convert to hours
Multiply by the average hourly cost of the employees using the tool

Example (round numbers only; use your actual figures):

8 hours saved per month
Average employee hourly cost: $45
Monthly value: $360
Monthly tool cost: $120
Monthly net gain: $240

If the monthly value exceeds the monthly cost, the tool is earning its place. If the calculation is close, check whether you are capturing all usage. Some tools are used more broadly than the initial tracking window captured.

Step 5: Account for Hidden Costs

Subscription price is not the full cost of an AI tool. Before making a keep-or-cut decision, add:

Onboarding and setup time (hours spent getting the tool configured and the team trained, multiplied by hourly cost)
Ongoing prompt and workflow maintenance (time spent refining how the tool is used each month)
Integration overhead (any manual steps the team has to take because the tool does not connect directly to the existing stack)

These costs reduce the net ROI. A tool with a low monthly fee but high maintenance overhead may be less valuable than the raw subscription math suggests. One that integrates cleanly and requires no upkeep has a higher realized value than the numbers alone show.

Step 6: Make the Keep-or-Cut Decision

At 90 days, you have enough data to make a confident decision for each tool. Three outcomes:

Keep: The tool consistently produces positive ROI, the team uses it regularly, and the tasks it handles would otherwise consume employee time.
Cut: The tool costs more than it saves, or adoption is low enough that the theoretical value is not being realized.
Replace or renegotiate: The tool is valuable but the pricing tier does not match actual usage. Downgrade to a lower tier or find a tool that does the same job at a lower cost.

Set a calendar reminder to run this evaluation every six months for tools you keep. Pricing tiers change, team workflows shift, and a tool that was a clear win in January may be underperforming by July.

Verify

Your measurement process is working if:

Every AI tool in the budget has a documented baseline and at least 30 days of time-tracking data
You can state the monthly ROI for each tool with a specific number, not an impression
The team doing the tracking agrees the data reflects actual usage (not ideal usage)
A keep-or-cut decision has been made for every tool at the 90-day mark

Troubleshooting

Users are not filling in the tracking log. The most common cause is that logging feels like extra work on top of the task. Reduce friction: a two-field Slack message ("task + minutes saved") is easier to fill in than a spreadsheet. If a Slack message bot or a form with autofill is available, use it.

The time savings are real but the ROI calculation is still negative. Check whether the tool is being used for the right tasks. AI writing assistants save the most time on drafts, summaries, and templated outputs, not on tasks requiring original analysis. If the tool is being used for low-value tasks, redirect it before cutting it.

Different team members report very different time savings for the same tool. Usage skill varies. A team member who has built effective prompts for a tool will save more time than one using default settings. Before cutting a tool based on low average savings, check whether the high-usage members could train the rest of the team. Often the gap is workflow knowledge, not tool quality.

Ready to Audit Your AI Stack?

ScaleIt helps startups and SMBs evaluate their AI tooling, identify what is earning its place, and configure the tools that stay to deliver consistent value. Book a free call to talk through what your current stack should look like.

Cross-referenced against IBM Think Insights AI ROI guidance (ibm.com/think/insights/ai-roi) and NCS London SMB AI ROI frameworks (ncs-london.com) on 2026-07-02.