For 15+ years, B2B leaders with per seat models treated DAU/WAU/MAU as something B2C companies obsessed over. We had ARR. We had NRR. We had logo retention. Engagement was a “nice to have” buried somewhere on the customer success dashboard, usually phrased as “utilization” that no one tracked for real. And hid from if it was zero.
With agentic apps, that era is over.
In B2B + AI, daily, weekly and monthly active users are now the single most predictive lighthouse metric you have. More than ARR growth. More than NPS. More than seat expansion. In the agent era, engagement isn’t a vanity number. It’s the leading indicator of every other number you care about: renewal, expansion, churn, even valuation multiple.
The CEO at Harvey just shared great metrics showing how usage is arguably the most important metric now in AI + B2B. Versus one we sort of kind of ignored pre-AI, outside of those with consumption-based models:
What Harvey’s CEO Posted
Three numbers from Harvey’s April:
- Net new ARR up 6x year-over-year
- DAU/MAU about to break 50%
- Average user spending 12 hours a month using Harvey
Those numbers belong together, and most B2B leaders still treat them as separate stories.
A B2B app with 50% DAU/MAU is statistically rare. Slack has it. Notion has it for power users. Most “successful” B2B tools sit at 10-20%. The S-1s of public B2B companies historically didn’t even disclose this number, because it was usually embarrassing.
12 hours a month per user is the more remarkable stat. Roughly 25-30 minutes a day, every working day, in a single vendor’s product. For comparison, the average ChatGPT session in 2025 was about 13-14 minutes. Harvey users are spending the better part of an hour per workday inside legal AI. In some ways, it’s no surprise give what it does for lawyers. But it also means Harvey is a workspace, not a tool.
Then look at the Net New ARR number. 6x year-over-year. Harvey crossed $190M ARR in January 2026 and just raised at an $11B valuation in March. They’re not growing 6x because of marketing. They’re growing 6x because their customers can’t stop using the product, and that engagement converts directly into seat expansion, departmental rollout, and firm-wide deployment.
DAU/MAU went up. Hours/MAU went up. Queries/MAU went up. Net new ARR followed.
In B2B + AI, engagement is the leading indicator. ARR is the trailing confirmation.
Why This Wasn’t True in So Much of Pre-AI B2B
In traditional B2B, you could be a dead product for two years before anyone noticed especially on seat-based pricing plans.
You sold an annual contract, sometimes multi-year. The buyer was a VP. The user could be an analyst three layers down. As long as your champion still had a job and the renewal got auto-renewed in procurement, you were “retained.”
Engagement didn’t really matter to the math. You could have a customer with 4% DAU/MAU paying you $200K a year and your CS dashboard would mark them green. That worked because there was nothing better. The switching cost of ripping out an enterprise system was higher than the cost of barely using it.
AI is changing all three legs of that.
The replacement costs are dropping to near-zero. An LLM can often dramatically simple migrations. In some cases, a team can even spin up a replacement workflow in Replit, Lovable or Cursor in an afternoon. The “we could never rip this out” moat is fading for any tool that isn’t deeply embedded.
Top AI-native software is also creating a higher engagement ceiling. Harvey users at 12 hours a month. ChatGPT Enterprise users at multiple sessions a day. Cursor users living in the IDE. The engagement bar in B2B is now being set by products people actively want to use, not products they have to use. If your product is in the second category, your customers are auditioning a replacement right now.
Buyers finally started measuring it. A recent Redpoint CIO survey showed 54% of CIOs are consolidating vendors and 45% of all new AI budget is being stolen from existing line items. The first thing CIOs cut is low-engagement tools. They pull up a dashboard, sort by DAU per seat, and the bottom of that list is the kill list.
Low engagement used to be a CS problem. Now it’s an existential one.
The Flip Side: Stealth Churn
Inside SaaStr, the other day we realized we hadn’t logged into Notion in months. We used to run our standup there. Then 10K (our AI VP of Marketing) became the dashboard. Then 10K became the standup. We didn’t decide to leave Notion. We just stopped opening it because our AI agents don’t need it.
Same thing happened with Canva. I used to use it almost every day, but I haven’t opened it in 100+ days. We replaced it piece by piece with Reve, Opus Pro, Higgsfield, OpenAI images and a few other AI-native specialists.
Stealth churn doesn’t show up in ARR. Yet. It doesn’t show up in logo retention. It shows up in DAU/WAU/MAU 6 to 18 months before the customer cancels.
Harvey at 50% DAU/MAU is close to the engagement ceiling. Notion inside SaaStr at 0% is the floor.
The Specific Metrics to Track
If you run a B2B + AI product, these are the numbers that should be on the wall. Not buried in a quarterly review deck. On the wall.
DAU/MAU ratio. The single best engagement number you can track. Below 20% means most of your users are casual. Above 40% means you have a daily habit product. Above 50% means you’ve built something people genuinely depend on. Track this monthly, by cohort, and by customer segment.
Hours per MAU. Harvey’s 12 hours a month is a remarkable number. Most B2B tools are at 30 minutes to 2 hours. If your product can credibly track time-in-product, this is the most direct measure of how much of the workday you own. If you can’t track it, get the instrumentation in immediately.
Queries or actions per MAU. Harvey’s queries/MAU jumped from ~60 to ~95+ in three months. The AI version of “feature engagement.” For agent-era products, this is more meaningful than session count, because one query can replace what used to be a 30-minute workflow.
Stealth churn cohorts. Customers paying you who haven’t logged in for 30, 60, 90 days. If you don’t have this dashboard, you don’t actually know your churn rate. You know your cancellation rate. Those are different numbers now.
Power user concentration. What percentage of your usage comes from your top 10% of users? In a healthy B2B + AI product, that number drops over time as more users find the daily habit. In a dying product, it rises as casual users churn out and only the diehards remain.
Make It A Top KPI
First: do you actually know your DAU/WAU/MAU ratios by customer?
Not aggregate. By customer. Aggregate hides the stealth churners.
Second: when usage drops 30% in 30 days, does anyone in your company know? Does … everyone?
Not your CS team during their quarterly review (which often is just focused on upsells these days). Today. If your customer’s engagement just collapsed, you have somewhere between 60 and 180 days before they cancel, and that window is the only chance you have to save them.
The B2B + AI companies winning in 2026 are running their enterprise products like consumer apps. At least in terms of usage.
They watch engagement daily, triage drops within hours, and treat DAU/WAU/MAU as the primary KPI with ARR as the lagging confirmation.
Harvey shows everyone what the top of the mountain looks like. 50% DAU/MAU. 12 hours a month per user. 6x net new ARR. That’s the difference between a B2B company that compounds and a B2B company that gets quietly replaced over the next 18 months.
Want to see what 50% DAU/MAU looks like inside the leading B2B + AI products? Come to SaaStr AI Annual, May 12-14 in the SF Bay Area. Replit, Salesforce, Google Cloud, Artisan, and 100+ B2B + AI leaders (including Harvey) going deep on the products their customers can’t stop using. Tickets at saastrannual2026.com.
