At Bessemer, we witness a deep shift that re-changes the landscape of SaaS. Not only does it stack up AI capabilities, it is also fundamentally rethinking the entire business to establish itself as an AI native. But the tactical blueprint for this transformation is often elusive.
Intercom stands out as a rare example of clarity in this shift and navigating with speed. Before the generation AI became mainstream, we were already experimenting with AI-driven customer support in 2018. In less than two years, the customer support leader has transformed AI agent FIN into a breakout AI company in order for AI agent FIN to win a $100 million AR.
Understanding how to migrate from Legacy Saas to AI-First Enterprise is a critical challenge and opportunity in our time. So we hosted Intercom Engineering SVP Jordan Neal to discuss the lessons from their journey and gather the most important takeaways for founders and executives who will lead their own AI transformation.
Main takeout from Intercom's playbook for AI founders and engineering leaders:
- Speed and depth design – Centralize AI talent. Organize mission-critical workstreams with clear ownership.
- Reorganize your foundations – AI is not an add-on. Scaling your codebase ai-native faster and smarter.
- Build ahead of the curve and prove your preparation – Internally prototype, mercilessly validate, open early, shape the future with real users.
- Impact price – Charge for results, not seats. Move from vendor to true partner.
- Hire adaptable builders – Empower product-oriented generalists and design engineers. Back them where it matters.
1. Redesign the operating model by centralizing R&D with a single read ownership.
Intercom's transformation began with a major top-down commitment. A few days after ChatGpt's release, the CEO announced that Intercom will become an AI-first company. This critical move has revealed that the company's traditional functionally siloed organizational structure is no longer suitable for purpose. To accommodate the rapidly evolving landscape, Intercom redesigned its behavioral model around two key principles.
1. A centralized shock team
The company focused key features, especially the AI team. The group expanded rapidly, growing from less than 10 to 50 machine learning (ML) researchers and scientists. Rather than spreading these experts across product teams, Intercom put them together to nurture a deep culture of experimentation and learning. Jordan Neal explained that the AI Group's success was measured by “what did you send this week” and “what did you learn this week?” This approach allows the team to release constant transport pressures and avoid local maxima traps and pursue fundamental advances.
https://www.youtube.com/watch?v=1i6k2wxzd_u
2. Mission-driven, sensual work stream
To perform speed, Intercom introduced “Workstream.” It's a small, dedicated startup-like team of 10-15 people drawn from engineering, ML, sales, marketing and more. Each workstream is assigned a direct manager (DRI), the single owner of any discipline that is given autonomy to drive a particular project for weeks or months. This replaces the spread of ownership and ensures focus. Today, Jordan says, “Nearly everybody at Intercom's R&D is working on FIN.” It's a fierce model, but what they believe is essential to catering to the market.
2. Re-depicting the codebase for AI-Native development – it is existential and treats it rather than an option.
A frequent objection from established companies is that they are simply too busy to reorganize the AI codebase. Intercom's experience reveals it: this is a false economy. To unlock the true potential of AI, it's not just a layer on top of a legacy system. The foundations need to be fundamentally reconstructed to AI natives.
Intercom's CTO has launched a bold “2x initiative” aimed at double the R&D output measured in shipped pull requests. This isn't just about adopting tools like Github Copilot and Cursor. It called for deep structural changes in both the technology stack and the development workflow. For example, the team is in the process of migrating the entire frontend from Ember.js for a response, not to chase the trend, but because, as Jordan said, “AI is way better than Ember code.” The team is also adapting backend systems and workflows to maximize the leverage that AI can deliver.
https://www.youtube.com/watch?v=gyksqqibwno
The move to overhaul the core parts of the product highlights the most important and perhaps the most difficult lesson for a company. It's trap to be too busy to tackle technical debt. Intercom proves that rebuilding unlocks efficiency later. There is inherent value in slowing down if it helps you gain accelerated speed later. Investing in basic changes to create a Tech Stack AI-Native is not a distraction from the roadmap. It's a roadmap to stay competitive. If development speed is a key determinant of success, then the codebase where AI efficiently reads, writes, and refactors are responsible rather than assets.
Intercom deploys an autonomous coding agent that proactively sends pull requests for tasks, such as removing the Dead Feature flag. The task “is similar to the same practices we do to coordinate our code for new hires,” Jordan points out. The key difference is that the leverage gained from AI is much higher, making the ROI of this fundamental work undeniable.
3. Develop before the function and gas is applied when ready.
In a rapidly evolving landscape, building for today's AI is a losing strategy. Intercom's success with FIN was driven by a future-thinking approach, and while developing ahead of the technology curve, it rigorously examined its preparation before exposing customers to new features.
https://www.youtube.com/watch?v=rf88d8fqwmo
First, build a “Taste Tester.”
Intercom built a sophisticated internal assessment framework before FIN became a customer-facing experience. This “machine for building machines” included backtesting against historical data, simulation of user behavior, and conducting large-scale A/B tests, verifying all changes to key metrics such as resolution, customer satisfaction, and hallucination frequency. Such a rigorous process proved that the GPT-4 was mature enough to launch, and was able to become a launch partner with confidence.
https://www.youtube.com/watch?v=gxqevmncll8
Risk using an internal prototype.
Internal prototyping is driving Intercom ahead of the curve without exposing customers to unstable technology. To evaluate cutting-edge models and ideas, Jordan emphasized, “We build it ourselves and experiment. It becomes very obvious when technology is not ready.” Their own support teams are often the first Alpha customers and provide a strong litmus test. If your team doesn't trust AI in their workflow, it shouldn't be just live.
“Publicate” to attract early adopters and shape the product.
When the prototype is internally promising, Intercom moves to a “publicly available” strategy for rapid market feedback. For example, they adapted the internal Friday show and tel sessions of the Fin.ai blog to public video content. This is not about releasing half-baked products, but it creates signal direction, excitement, and attracts motivated groups of early adopters and design partners, helping shape the future of your product, even if it's still a work in progress.
4. Adjust prices and value to transform commercial relationships.
AI isn't just about rebuilding products. Redefine how you create, deliver, and capture customer value. Intercom's transformation into AI-Native business coincided with a fundamental overhaul of the commercial model by ensuring pricing, incentives and customer success were vigorously aligned. The company did this in two important ways.
1. Not tools, but results pricing.
Intercom defeated the traditional SaaS Playbook by pioneering FIN results-based models. Instead of charging per user or sheet, the customer only pays if the FIN autonomously resolves the conversation. As Jordan pointed out, “If Finn has to escalate to a human, you won't pay.” This was more than a price adjustment. This was a strategic lever that changed internal behavior and customer relationships. With results-based pricing, all team members from product to sales now have North Star, resolved conversations. Pricing models based on what AI can offer have impacted a wide range of industries that employ AI, with competitors like Salesforce and Zendesk unveiling similar models after FIN.
2. Transition from sales to partnersing.
The move to outcome-based pricing required an intercom to rethink the GTM strategy. Jordan compared the traditional SaaS sales strategy to “car sales.” This can be a transaction exchange that leaves a lot of risk in the hands of the customer. For this reason, Intercom has introduced future deployment engineers who work with customers to ensure successful setup and recruitment. This approach recognizes that deploying AI is not something to flip a switch. It's about guiding customers through change management.
https://www.youtube.com/watch?v=t25pssuc_oy
5. Hire for versatility and empower generalists and design engineers.
The evolution towards AI-Native's work has changed the meaning of building a great team. It places premiums on multipurpose, product-oriented individuals who can actively break down strict silos between disciplines and own end-to-end issues. Intercom's employment philosophy aims to find and empower this talent through three key principles.
1. Prioritize product-oriented generalists – Intercom has long been at the heart of recruiting “product engineers” who are not only technologically keen, but also deeply motivated by their impact on customers and business. In this new paradigm, these generalists are thriving. They are empowered to work directly with their customers, identify problems and provide directional solutions.
2. Accepting “design engineers” – As we discovered from unlocking the evolution of full stack design, the emergence of “design engineers” is bringing tactical change. At Intercom, designers are no longer creating mockups, they are also creating and committing production code. This hybrid role eliminates handoffs and ticket bottlenecks, allowing designers to fix themselves as soon as they find them.
3. A balance between generalists and deep experts – The Intercom product team is built around a versatile generalist, but the Core AI Group is intentionally specialized. The company is looking for a specific profile. MLPhD researchers and scientists have also experienced large-scale shipping products. This double approach – strengthening the broad foundation of generalists while burning the central brain of world-class professional trust ensures advances in basic technology.
The patterns across all the changes on Intercom are clear. Successful AI conversion requires short-term disruption for long-term progress. Their experience shows that overall thriving SaaS leaders are people who are willing to fundamentally rethink their business for AI.
Subscribe to Atlas for detailed insights into building AI-powered products. Also, if you are undergoing similar transformations, we look forward to hearing from you. Please contact bnagda@bvp.com.
