We often hear narratives about AI implementation that sound neat and linear, but the reality inside a product team is usually much messier.
This is a look at how Nodalview navigated that complexity, based on the experience of their Product Manager, Sam Boribon. They stumbled twice, not because they lacked the technology, but because of how they organised their teams before finally figuring out how to treat AI as a foundation rather than just a feature.
The starting point
Nodalview is a real estate marketing platform designed to enable agents to generate visual assets, including photos, videos, and floor plans, directly from mobile devices. Established in 2016, the company has progressed through the typical SaaS lifecycle phases: bootstrapping, venture-backed scaling, and a return to profitability. This context is critical, as every decision regarding AI was influenced by the company's financial, organisational, and strategic standing at the time.
It is important to note that AI was not new to the Nodalview product. Long before the rise of generative AI, Nodalview already relied on computer vision to enhance images, handling tasks such as perspective correction, noise reduction, and brightness optimisation. AI was already embedded in the product, though not yet in the generative form the market would later demand.
Phase 1 - internal capability building
As generative AI gained market traction and AI-native competitors emerged, Nodalview’s first instinct was to build the capability in-house. They hired a strong academic expert in computer vision, which seemed like the right move to build a solid, long-term technical foundation.
In practice, however, progress was slow and fragile. The issue wasn't the expert's skills, but the fact that she was working in isolation. The expert lacked a team to challenge her ideas or a feedback loop with the product to iterate quickly. While she focused on theoretical robustness, the market was moving fast, and the organisation wasn't set up to absorb or scale what she was building resulting in an "AI vision" that was practically non-existent.
Validating demand with off-the-shelf solutions
Concurrently, Nodalview tried integrating off-the-shelf AI models to deliver specific features, like the "Magic Eraser" to remove unwanted elements from photos. This approach successfully validated market demand, confirming a genuine user appetite for AI-powered tools. Users readily adopted these features, proving that AI could solve real pain points in the real estate marketing workflow.
However, this phase also highlighted commercial and operational constraints:
- Unit economics: Unlike traditional SaaS where the marginal cost of a new user is near zero, AI costs scale linearly with usage. To fix the margins, Nodalview had to introduce "AI Credits," which was initially a nightmare for their customers, breaking their mental model of a flat subscription.
- Lack of control: Nodalview had limited control over the performance and evolution of the models. If the third party provider changed the model or pricing, Nodalview was exposed.
- Strategic limitation: At this stage, AI remained an add-on valuable for specific tasks, but not a structural pillar of the product. The company was reacting to feature requests rather than driving a strategy.
Phase 2 - acceleration via external expertise
In a following effort to accelerate development, Nodalview partnered with external AI specialists to deliver a specific outcome: automated home staging to realistically furnish empty rooms.
The initiative was well intentioned and planned, with defined outcome and detailed requirements. However, the results failed to meet production standards. Retrospectively, the failure was due to a misalignment between internal readiness and the complexity of applying the solution in a real world context.
While this phase accelerated internal learning and deepened AI understanding, it came at a significant financial and organisational cost, with AI still responding to product strategy rather than driving it.
The structural gap
Looking across the first two phases, a clear pattern emerges. Nodalview did not fail because of poor technology choices, insufficient talent, or weak partners. They failed because AI was consistently treated as a response to product needs, rather than as a capability that reshapes how product decisions are made.
In phase one, AI expertise was isolated. In phase two, it was outsourced. In both cases, the organisation lacked the structures required to absorb and evolve what was being built. AI existed either as an individual effort or as an external deliverable, but never as a shared responsibility across product, design, and engineering.
This is a common inflection point. At this stage, organisations often believe the problem is speed or execution, when in reality, it is sequencing. AI is approached as something to implement, when it should first be framed as something the organisation must learn to carry.
At dualoop, this is typically where we intervene to reset how AI initiatives are framed. We start by clarifying which user and business outcomes AI is expected to move, establishing economic and trust constraints early, and distributing ownership across teams so AI becomes a collective product capability rather than a specialist dependency.
Without this work, AI remains impressive but fragile. With it, AI can become a foundation rather than a feature.
Phase 3 - the turning point: acquiring living capability
What finally worked for Nodalview was a structural shift. They acquired two companies: Flash, a small team deep into generative AI, and Proper Shot, a competitor.
The acquisition of Flash was the pivot point: they were not just buying code but were acquiring a team that already had a rhythm of shipping, iterating, and learning from real users under constraints.
Context was again a key factor: Nodalview was profitable and trust had been established with the founders over time. Although complex, the acquisition aligned with the broader business strategy.
The cultural shift behind AI success
The acquisition would have failed if Nodalview had not focused on the human side of the integration. Teams were brought together physically, and time was invested in team-building activities and shared ways of working to improve collaboration and shared ownership. Organisational structures were also updated to make AI capability a core part of Nodalview, rather than something managed at the edges of the organisation.
This change was essential as AI systems evolve quickly. Without a team continuously improving and maintaining them, their value quickly declines. AI moved from being a set of individual features to becoming a foundational part of the product.
They operationalised this by turning their user base into their QA team. Immediately after an agent generates an image, a simple pop-up asks if they are happy with the result. These responses pipeline directly into a Slack channel visible to the engineers, creating an immediate loop where developers can catch model hallucinations in real-time without needing a large internal QA department.
How can organisations grow AI capabilities organically?
For organisations where acquisition is not a viable option, the lessons from Nodalview still apply:
- Validate demand: Begin by integrating off-the-shelf solutions to confirm user needs before committing to custom development.
- Empowerment: Engage AI experts as coaches and enablers, to upskill your team rather than asking them to build everything in a black box.
- Agility: Maintain alertness to cutting edge solutions, as the landscape evolves faster than internal roadmaps.
- Flexibility: Remain model agnostic to avoid vendor lock in
Conclusion
This maturity has allowed Nodalview to shift their horizon, by expanding beyond their own application and starting to act as infrastructure for the rest of the industry. By packaging their AI tools into an SDK, they are now embedding their capabilities directly into external real estate portals and CRMs, marking a distinct evolution from building tools for agents to becoming the engine that runs other platforms.
Ultimately, AI should be treated as an organisational and product decision, not just a technical one. Nodalview’s initial setbacks occurred because AI was treated as an optimisation problem rather than a structural one. Success was achieved not because of superior talent or exclusive technology access, but because the company built a strong collective capability around the technology. Once AI became a foundation rather than an add-on, the outcomes fundamentally changed.
