In partnership with
AI Agents Are Reading Your Docs. Are You Ready?
Last month, 48% of visitors to documentation sites across Mintlify were AI agents, not humans.
Claude Code, Cursor, and other coding agents are becoming the actual customers reading your docs. And they read everything.
This changes what good documentation means. Humans skim and forgive gaps. Agents methodically check every endpoint, read every guide, and compare you against alternatives with zero fatigue.
Your docs aren't just helping users anymore. They're your product's first interview with the machines deciding whether to recommend you.
That means: clear schema markup so agents can parse your content, real benchmarks instead of marketing fluff, open endpoints agents can actually test, and honest comparisons that emphasize strengths without hype.
Mintlify powers documentation for over 20,000 companies, reaching 100M+ people every year. We just raised a $45M Series B led by @a16z and @SalesforceVC to build the knowledge layer for the agent era.
There is a version of AI adoption that feels productive without actually being productive. You are using more tools. Your team is doing prompting. Things feel more modern. But when someone asks you what the ROI is on all of this AI spend, you go quiet.
That is the uncomfortable truth sitting behind a statistic from yesterday's roundup: only 29 percent of executives say they can measure AI ROI confidently. The other 71 percent are somewhere between 'we think it is helping' and 'I genuinely have no idea.'
This is the Sunday Strategy edition. By the time you finish reading this, you will have a framework you can start applying tomorrow morning. No data science degree required. No expensive analytics stack needed. Just a clear-headed approach to measuring what actually matters.
Why AI ROI Is Hard to Measure
Before we get into the how, it is worth understanding why most businesses struggle with this. There are three culprits.
First, AI is often bundled with other changes. You automate a workflow and your response time improves, but you also hired a new person around the same time. Which caused the improvement? This isolation problem is real and annoying, but it is solvable with the right setup.
Second, gains are often indirect. AI saves time, but time savings do not automatically show up in your P&L. Someone works faster, but do they generate more revenue, or do they just respond to more Slack messages? The connection between time saved and money made requires deliberate tracking.
Third, nobody established a baseline before they started. You cannot prove improvement without knowing where you started. This is the most common and most avoidable mistake. The fix costs you about 20 minutes of work before you launch anything.
The Framework: Four Categories of AI Return
AI ROI does not fit neatly into a single formula. It shows up in four different categories, each requiring different measurement approaches and different timelines.
Category 1: Cost Reduction
This is the easiest to calculate and the fastest to materialize. Cost reduction ROI comes from AI handling work that previously required paid human time. Think automated lead follow-up replacing manual outreach, AI-generated first drafts reducing copywriting hours, or automated invoice processing replacing administrative time.
How to measure it: Track hours saved per week on the specific task before automation. Multiply by your internal cost per hour for that work. That is your weekly cost reduction. Annualize it. Compare to the cost of the tool.
Example: If you spend 8 hours a week on lead follow-up at an effective cost of 50 dollars per hour, that is 400 dollars per week. An automation that handles 80 percent of that work saves you 320 dollars per week, or roughly 16,640 dollars per year. If the tool costs 100 dollars per month, the math is not complicated.
Category 2: Revenue Acceleration
This is harder to isolate but often delivers the largest numbers. Revenue acceleration comes from AI enabling faster response times, more personalized outreach, better content, or more consistent follow-up, all of which move deals faster and close more of them.
How to measure it: Pick one revenue metric you can attribute specifically to an AI-assisted process. Lead-to-meeting conversion rate is a clean one if you are automating follow-up. Average deal size if you are using AI for proposal personalization. Track that metric for the 90 days before you implement and the 90 days after. The delta is attributable to the change.
This is where the baseline setup matters most. Before you launch any AI-assisted sales or marketing workflow, write down your current conversion rate, your current average response time, and your current close rate. You will thank yourself in three months.
Category 3: Risk Mitigation
This is the category most businesses ignore entirely because it is the hardest to put a number on. But AI-powered monitoring, whether for cash flow anomalies, customer churn signals, or support ticket trends, is preventing problems that would otherwise cost you real money.
How to measure it: Track instances where an automated alert led to a proactive action that averted a problem. Estimate the cost of that problem going undetected. Customer churn prevention is a useful example. If your AI monitoring flags a customer showing disengagement patterns and your team intervenes before they cancel, the value is the lifetime revenue of that retained customer minus the cost of the intervention.
This is imprecise by nature. But even a rough estimate beats treating this category as zero, which is what most businesses effectively do.
Category 4: Capability Expansion
The fourth category is what happens when AI lets you do things you simply could not do before. Not faster or cheaper versions of existing work, but genuinely new capabilities. Publishing daily content when you previously published weekly. Personalizing outreach at scale when you previously sent batch emails. Running 24/7 customer support when you previously covered business hours only.
How to measure it: Identify what the equivalent cost would be to achieve this capability through traditional means. Hiring a content writer for daily publication. Staffing overnight support. Building a personalization team. The gap between that cost and your AI tool cost is the efficiency gain.
The 20-Minute Baseline Setup
Here is what to do before you implement any new AI workflow. It takes 20 minutes and it is the single most important thing you can do to make your ROI measurement work later.
Pick the metric most directly affected by the workflow you are building. One metric. Not five.
Pull the last 90 days of data for that metric. Write it down somewhere you will not lose it.
Note the current cost in time or money of the process you are automating.
Set a calendar reminder for 90 days from your launch date to run the same measurement.
On that date, compare the numbers and calculate the delta.
That is it. Five steps, 20 minutes of setup, and you have a measurement framework that will tell you something real in three months. Most businesses skip this entirely and then wonder why they cannot answer the ROI question with any confidence.
The Quarterly AI Audit
Once you have active workflows running, build a quarterly audit habit. Four questions, answered with actual data, every quarter.
Which AI workflow delivered the clearest measurable return this quarter?
Which workflow is running but has no clear measurement attached to it?
Are there workflows I set up and essentially forgot about that may or may not be running correctly?
What is the one new workflow that would deliver the highest measurable return next quarter?
The audit is not about justifying your AI spend to a board. It is about making better decisions with your own money. The businesses winning with AI right now are not the ones spending the most. They are the ones learning the fastest. And you cannot learn without measuring. Tools like Fathom for meeting analytics and Make.com for workflow logging give you the baseline infrastructure for this kind of visibility. They tell you what is running, what is working, and where the gaps are.
The Mindset Shift That Makes This Work
Here is the last thing worth saying on this topic, and it is the thing that actually separates businesses that get ROI from AI from the ones that keep talking about it.
Stop treating AI as a technology decision and start treating it as a business decision. Every tool you adopt should have a measurable job. Every workflow you build should have a number attached to it. Every quarter you should know, specifically and defensibly, what your AI stack returned on the investment.
When you approach it this way, two things happen. You make better decisions about where to invest and where to cut. And you stop being intimidated by the ROI question because you actually know the answer.
The businesses that are going to look smart in two years are not the ones who were earliest to try AI. They are the ones who got disciplined about measuring it. That discipline is available to any business owner willing to spend 20 minutes on a baseline before they launch the next workflow.
Start there. Everything else follows.
Ready to build a fully measured AI workflow system? Reply ACCELERATOR to get the AI Business Accelerator ($97) which includes our complete ROI tracking templates, quarterly audit framework, and the workflow blueprints that have the clearest measurement paths for small businesses.