The future of Product-led Growth with AI powered CRMs
A case study on how disruption feels
Sponsored by Attio, the CRM for the AI era.
Sync your email and watch Attio build a powerful CRM - with every interaction you’ve ever had, totally enriched and organized.
👉 Start your free trial today.
In 2020, we had over 100,000 paying customers at Smallpdf. We also had the understanding that we were a pure B2C product since we only had a handful of messages from B2B clients through our support per week.
I ran some small experiments at the time to see what would happen if we made it easier for B2B clients to send us requests for simple things like bulk discounts and shared logins.
Turns out we were not as B2C as we thought: businesses started to contact us in bigger numbers, even though we had almost nothing to offer to them specifically.
Well, we needed a process, CRM, custom billing, and people to manage it.
I tried and brought Hubspot to Smallpdf to sunset our crappy excel sheet which did all the heavy lifting at the time within a couple days.
It did do its job well and was such an improvement over what I knew from my Microsoft Dynamics and Salesforce days.
The topic of “CRM at Smallpdf” was kinda “done” in my mind at that point.
But 2020 is long gone, and things have changed dramatically in the realm of CRM, and it’s a case study for what’s happening to a lot of other verticals as well:
Product-led growth (A freemium + slick onboarding) is not special anymore, but the bare minimum.
What we need to “store” about our future customers has changed
It’s not salespeople that interact with CRMs anymore, it’s entire organizations.
Let’s dig into it.
Polishing a CRM turd
In almost every company I was involved with in the past, the same thing happened sooner or later:
When I started to look at the customer experience for B2B clients, I found that interacting with sales was a total mess …having potential.
The solution is usually:
“We have to start bringing behavior signals to the attention of our salespeople so they can close better.”
For instance, whether an account had reached its aha moment. Or which part of the application they have used most before calling us.
We wanted to keep our salespeople informed, and the customer from telling us stuff we already knew anyway when they picked up the phone.
And every time I wanted to enable that, I had to bend these tools to my will by creating flags or weird feeling data streams, which broke over time and took time to implement.
It was acceptable 5 years ago and still worth doing because it meant a competitive advantage once we were done.
Today, it just feels like a churn reason, I’m sick of writing PRDs to facilitate it, and I sometimes even explore with my clients whether they are open to ditching their CRM tool in a painful process.
Rather than working around these systems, I’d like them to be built for that.
But this is not just a little feature request… It’s a reflection of what has changed:
Traditional CRMs, built for sales-led motions, struggle to capture the real-time product usage data that fuels PLG strategies.
Meanwhile, AI-native CRMs are coming up as the solution, blending customer insights, predictive analytics, and automation to create hyper-personalized experiences at scale.
In other words, new tools that win are crossfunctional first, and silo driven second.
1. Why Traditional CRMs Fail in a PLG World
Obviously, the traditional CRM companies are not blind to this, but they struggle to pivot because they have so much pull from their existing customers.
This is not a problem that they can optimize by making the interface a little slicker and easier to use. This goes to the core of their offering. If they change, they lose those existing customers who are not ready to switch yet.
The core issues they struggle to address:
Blindspots in free/user tiers: Legacy systems ignore product-qualified leads (PQLs) from free trials or freemium models, missing critical signals like feature adoption or engagement spikes.
Reactive vs. proactive: Manual data entry and static pipelines can’t keep up with real-time user behavior, delaying interventions for at-risk accounts or upsell opportunities.
One-size-fits-all outreach: Without AI-driven segmentation, campaigns lack the precision needed to convert self-serve users into paying customers. You cannot simply turn up the noise, you want your customers to listen when you speak to them. That’s only possible if you contact them when it’s relevant to them. (In channels that matter to them, too)
2. How AI Closes the PLG-CRM Gap: Timing
AI transforms CRMs from passive databases into active growth partners. The following three gaps are all doable with traditional CRMs, but none of them do them well:
Predictive lead scoring: Analyzes product usage patterns (e.g., feature adoption frequency, session duration) to identify high-intent PQAs and PQLs. Lead scoring per se is nothing new, of course, but it usually stops on firmographics (how big is a company, location, etc.) with most systems. This does not inform you at all on when to reach out. (Behaviour does)
Automated workflows: Triggers personalized nudges (e.g., in-app messages, tailored demos) when users hit predefined engagement thresholds. A well-defined threshold (commonly a measurable AHA moment) is extremely powerful for your salespeople. These prospects will not only answer your outreach, but they will also have questions and appreciate a consultative approach right at that moment.
Churn prevention: Flags accounts with declining activity and recommends targeted re-engagement campaigns or interventions. This is similar to the lead scoring above, but with churn and for customer success.
Traditional CRM’s try to do this on a good day with the NPS, which is absolutely useless. Objective user engagement thresholds are warning you well in advance, sometimes months before they will cancel.
3. Implementing AI-Driven PLG Processes: Tacticals
When you’re reading this, chances are you work in a company with CRM tooling from one of the major players like Salesforce, Hubspot, or Dynamics.
And realistically, most of my clients are also not willing to switch tools right away, but I’m seeing more and more signals that this is bound to happen with time.
At the very least, for newer companies, it’s worth it to evaluate some alternatives. If I wanted to future-proof a new B2B company and could choose something, this is what a great CRM tooling/org needs to cover:
Integrated cross-functional product analytics: Able to feed usage data (e.g., API calls, dashboard views) into your CRM to build 360° customer profiles. Most actionable product usage data is simple in nature. That’s why we create simple Aha Moments in the product. Make sales and customer success use them, by not drowning them in other noise.
Track your usage data, train AI on historical wins: Use past conversion data to identify high-value PQL patterns (e.g., teams inviting 3+ members in 7 days). If you don’t have this data yet, make sure you’re tracking it. You can always switch tools, but you won’t be able to track data retroactively.
Launch micro-campaigns: Test AI-generated email variants or manual outreach targeting users who abandon workflows mid-task after having intense engagement.
Measure what matters: Track metrics like PQL-to-paid rate and AI-driven upsell revenue instead of vanity lead counts and MQL conversions. We care less about “who” (sales or your product) closes a PQL, but how fast we can do it. While we can’t hurry our customers’ internal processes, we can optimize ours through our product and internal processes.
Transparent cross-functional scoring models: Getting one good, simple behaviour scoring is even with the correct tooling hard, but one key is to involve everyone.
Even for me, who has done this dozens of times, it’s never as simple as “engagement” = “they want to buy”. Whenever it gets complex and doesn’t involve people from product, sales, and customer success, it fails. Every time.
I don’t want tooling to get in the way of that, I want it to help me.
4. The Future: Autonomous CRMs and Frictionless Growth
Today's demands are emerging in that regard quite clearly for me, and I think many of the above thoughts can be applied to other verticals as well.
If I put my tinfoil hat on and look even further, I’d bet on the following trends and would structure my product strategy accordingly:
Auto-negotiate contracts: AI agents handle discount requests or renewal terms based on customer health scores, and not just current customer spend.
Hyper-personalized, hyper-relevant content that doesn’t imitate humans: I’m sick of AI tooling pretending to be human and faking a genuine connection. Get to the point and tell me only what’s relevant. If it is, I will pay attention.
I’m happy to listen to an AI outreach tool if what it tells me is actually relevant and a win-win for me.
That’s only possible with a great predictive model.Predict and action revenue leaks on medium-sized accounts: Proactively alert teams about at-risk accounts 30+ days before churn out of the box. This is already happening for high-value enterprise clients (through their Account executives), but the threshold where it’s profitable will start to come down, enabling it also for smaller and more accounts.
Summary: CRM is becoming a product-people problem
I’ve written on this topic mainly from the perspective of tooling, but I do believe that Customer Relationship Management in the future is becoming more and more an organizational product problem and not just a “let’s get a tool” problem.
In a recent talk I gave at Visma, I used this slide to summarize it:
Let’s look at the problems that future tools will solve in point 1 above:
Blindspots in free/user tiers
Reactive vs. proactive
One-size-fits-all outreach
You will realize that out-of-the-box tooling can only get you so far. In order to enable these future AI tools, product people need to enable them.
Whether it’s messaging, solving these blind spots for sales, or starting to be proactive, it requires the tools to have access to clean usage data and be able to interpret it correctly.
Making an org and it’s data compatible in this regard will be a core competence of a good product and growth leader, including understanding the cross-functional processes inside your company.
Future products are not just the sofware you’re building, but also the sales people and marketing messages and onboarding flows that interact with our customers. Start to understand and manage them like you do customers and no AI will ever replace you.
Sponsored by Attio, the CRM for the AI era.
Sync your email and watch Attio build a powerful CRM - with every interaction you’ve ever had, totally enriched and organized.
👉 Start your free trial today.
This is the most accurate take I’ve seen on the disconnect between traditional CRMs and what modern GTM teams actually need. Most tools still treat product usage like a “nice-to-have,” when it should be the foundation of every touchpoint.