Q3 AI review
Hi Everyone,
If you've added AI tools to your company over the past 18 months, some of them are probably sitting there unused.
A 2025 S&P Global survey of 1,006 IT and business leaders found that 42% of companies abandoned most of their AI projects before they were fully launched. A year earlier, that figure was 17%.
Before you spend more on AI this quarter, take a closer look at what you already pay for.
Here are five checks to run on every AI tool in your company.
Check 1: Has the tool changed how the work gets done?
This is the most important check.
McKinsey surveyed 1,993 organizations in mid-2025 and found that 88% were using AI in at least one part of the business. Only 6% said AI had increased earnings by at least 5%.
The companies seeing the strongest results had changed the work around the tool. They had removed steps, changed who handled each task, and found ways for AI to take on part of the process. Others had added a new tool while keeping the old process the same.
Ask the person closest to the work:
Has this tool changed how you complete the job?
An answer like "I still do everything the same way, but I also have this tool open" suggests the company has added another step rather than improved the work.
Check 2: Who is using it each week?
For every tool, compare two numbers:
- How many licenses are you paying for?
- How many people used the tool during the past week?
Ignore anyone who logged in once six months ago. You need to know whether the tool is part of their work today. A useful starting point is 30%.
When fewer than three in ten paid users are using a tool each week after the first three months, place it on the review list and agree on a final deadline.
Check 3: Do you already pay for something that does the same job?
This is often where the easiest savings are found.
Salesforce, HubSpot, Microsoft, Notion, and Zoom have added AI features to the products their customers already use. A separate tool may now do the same job as software you already pay for.
Review each tool and ask:
- Does it give us something we cannot already get elsewhere?
- Is it good enough to justify the extra cost?
- Would the team lose anything important if we removed it?
Check 4: What number should it improve?
Every tool should be connected to one clear result.
That could be time spent on each support ticket, calls handled by each sales representative, hours required to create a report, or time needed to complete a code review.
You should know what that number was before the tool arrived and what it is today.
A missing number leaves you without a clear way to judge the tool.
Check 5: Who is responsible for the result?
Every AI tool needs one person responsible for what it is meant to achieve.
Most tools enter a company because one person pushes for them. They find the product, make the case for buying it, and encourage the team to use it.
Then that person leaves, changes roles, or moves on to another project, and the tool remains in the budget.
Name the current owner of every AI tool in your company.
That person should be able to tell you what the tool should achieve this quarter, how you will measure the result, who should be using it, and what needs to happen before the next renewal.
Run this review before Q3
Block 90 minutes this or next week. Include Finance, IT, or security, and someone who works across departments.
Finance should bring the costs, contracts, and renewal dates. IT or security should bring the usage figures and a list of who has access to each tool. The person working across departments can help turn the review into decisions.
Make a list of every AI tool the company pays for and run each one through the five checks above. Then place each tool into one of three groups:
Keep: It is used regularly and produces a clear result.
Review: The value is still unclear, with a plan and deadline to improve it.
Remove: It has little use, repeats something you already own, lacks an owner, or creates a concern around company data.
Go deeper
👉 McKinsey: The state of AI in 2025 — what the 6% of companies seeing real returns from AI do differently
👉 S&P Global: Generative AI shows rapid growth but yields mixed results — why companies are walking away from many of their AI bets
👉 IBM: Cost of a Data Breach Report 2025 — how unapproved AI tools can add $670K to the average breach
👉 CIO.com: CIOs cut IT corners to manufacture budget for AI — how mid-market CIOs removed or combined software to fund AI
Coming up tomorrow
Tomorrow, we'll share a seven-question filter to help you manage your busy inbox.
P.S. How many AI tools is your company paying for right now? Write down your guess before you pull the full list.