.jpeg)
In this article, we sum up some of the key takeaways from the broadcast, which featured some extremely high-profile names.
If you’d like to catch up with any of the sessions below, they’re available to stream on demand here:
Bret Taylor, Chairman of OpenAI and co-founder of Sierra
We started the broadcast with a fireside chat recorded with Bret Taylor, former CTO of Facebook, Chairman of Twitter, co-CEO of Salesforce, current Chairman of OpenAI, and co-founder of Sierra.
Technology defines companies, not the other way around
If you look at the largest companies in the world, most of them were born out of big technological shifts. Amazon, Google, and Salesforce were all born of the internet.
Bret and his team are doing really important things at Sierra, but he keeps himself grounded in the fact that really what he’s doing is riding the larger wave of technological advancements that we’re seeing with LLMs and AI agents.
Bret advises business leaders to avoid taking a myopic view of the world, in which you think you’re the only factor defining your own future - you’re not. You exist in a world of important externalities that you should factor into your decisions.
Build around imperfections to ensure trust in AI systems
Bret spoke about how he initially found building on top of LLMs to be incredibly challenging at first, owing to their non deterministic nature. He had two pieces of advice:
- Don’t over promise. When an AI agent is in the wild, it should identify itself as such to a user. People are aware they have the capacity to make mistakes, and they’ll be understanding of that.
- Ensure you have guardrails in place. Having models make important decisions isn’t particularly responsible right now, so ensure you have human auditors checking the work of AI and finding ways to improve it.
Basically - don’t wait for AI to be perfect. Assume it isn’t, and work around it.
Mati Staniszewski, co-founder and CEO, ElevenLabs
Bret’s session was followed with a sit-down with Mati, co-founder of AI voice platform ElevenLabs.
People are your most important asset - organize them well
Mati spoke about how he and his cofounder hire people: they have a clear idea of what they think the future is going to look like, and then they prioritise hiring people whose vision is aligned with that picture.
Organisationally, ElevenLabs is split into lots of small teams, each with a very high degree of independence. This ensures everyone feels like they have a real impact and ElevenLabs are able to execute quickly.
And within every team, Mati ensures there’s technical engineering talent available to find ways of making processes work better and faster. This is a particularly important consideration in the age of AI, where every team is looking for ways to find efficiencies and innovations.
Olivier Pomel, co-founder and CEO, Datadog
Next up was Olivier Pomel, who heads up Datadog, a monitoring and analytics platform that provides full-stack observability for cloud-scale applications and infrastructure.
Combine a top-down approach to AI with a bottoms-up one
It’s good to have big initiatives that touch every part of the org - those need to be driven from the top down. If you’re embracing new AI-enabled collaboration software for example, that’ll involve teams across every function.
But at the same time, it's important to give room for all of the various teams and pockets in the organization to build and innovate. They know their problems better than anyone else, so you need to enable them to find ways of solving them - rather than enforcing a particular approach.
Prioritise customer feedback
Olivier spoke at length about how Datadog maintains such a good state of profitability. Early on, the team decided they wanted very fast and clean feedback loops on the product. For Datadog, the main feedback loop is revenue and retention: how many people are buying the product, and are they keeping it?
To ensure they get prompt feedback, Datadog sells monthly plans - which is not a particularly common approach in enterprise. Software can often be sold for three years at a time - but Olivier says that can be a bad thing, because ‘you won’t get bad news fast enough.’
Des Traynor, co-Founder & CSO, Intercom
Next up was Des Traynor from AI-driven customer service and messaging platform Intercom.
Explore every possibility
AI massively decreases the cost of exploration. At Intercom, that meant roadmaps could become something almost interactive. Instead of debating which direction to pursue, teams could vibe-code an entire set of ideas – mock up interfaces, generate flows, even build rough prototypes – and evaluate them side by side before committing.
Aside from product, this approach reshaped the way leadership made decisions, too. Des noted that most people readily accept that AI can make engineers more productive. What fewer leaders consider is how much it can improve their own judgment. “Can AI make a VP better? Absolutely,” he said. “But many leaders still think AI is for the kids.”
AI and the build-versus-buy trap
Companies dramatically underestimate what it takes to build AI tools that actually work. On the surface, it’s tempting. A quick demo, a hackathon prototype, or a simple model prompt can give the impression that an internal solution is only a few weekends away.
But most teams don’t grasp the depth of the problem they’re stepping into. The difference between a demo and a dependable system is the difference between a hobby project and a product like Intercom’s AI agent Fin, which relies on dozens of subsystems, custom post-training, guardrails, and constant architectural refinement.
That complexity raises a harder question. Even if a company could build it, should they? Des’s stance is clear. “The majority of your talent should focus on advancing your strategy,” he explained, “not replacing internal tools to shave a few grand off.”
Naveen Zutshi, CIO, Databricks
Next we spoke to Naveen, CIO at Databricks - a cloud-based data and AI company that provides a unified platform for data engineering, machine learning, and analytics. Here’s what he had to say…
Your data and your vision are equally important
Successful AI adoption is dependent on data maturity and the quality of the use case. Companies that succeed have clear, governed proprietary data and well-thought-out use cases with business buy-in.
Naveen gave a few examples that Databricks has been able to support:
- Mastercard: Using AI for credit card processing and customer onboarding with human oversight.
- Rolls-Royce: Fusing IoT and environmental data to predict and prevent maintenance outages on aircraft engines, which reduced carbon emissions.
- Walgreens: Optimizing logistics to ensure the right drugs get to the right pharmacies on time, leading to a 20% improvement in productivity and $160 million in additional value.
CIOs must focus on clarity and responsible governance
CIOs face high demand for AI solutions from all departments and seniorities. Without a proactive approach, this demand can lead to "shadow AI" solutions popping up across the organization, which risks data leakage, security issues, and unintended consequences.
To mitigate this, every CIO should maintain an AI security framework, define a clear policy for the use of AI tools, and create an internal AI strategy document. Responsible governance also includes looking for bias, ensuring data lineage, and having auditability and traceability to correct agent mistakes.
Panel session - Michael Miao, VP of Finance, Glean, and Dan Zhang, CFO, ClickUp
Pigment’s Jay Peir was joined by two finance function heavyweights to discuss scaling AI from concepts to results.
Be smart with how you drive adoption
Dan outlined a two-act approach to driving AI at ClickUp. Act 1 is for leaders to frame AI as an opportunity for employees to create and innovate, supported by rigorous operating cadences like weekly training, monthly awards, and quarterly hackathons. Act 2 is to map AI projects to key organizational capabilities, such as streamlining transactional activities and deepening business partnerships.
Michael emphasized that successful AI adoption is an organizational challenge, not just a technology one. Since AI tools can be intimidating, ClickUp believes in showing users the value they provide, doing on hand holding and relying on roles they call ‘AI outcomes managers’ (similar to a forward-deployed engineer concept) to partner with departments and help them build and use AI agents.
Measuring return on investment requires a broad framework
Dan, advised breaking AI ROI into three distinct categories. The first is ‘1-10 automation,’ which involves straightforward, measurable time saved from eliminating manual tasks. The second is ‘0-1 unlocks,’ which refers to new capabilities that were impossible before AI, such as instantaneously customizing insights for every user or scanning every community comment.
The third category focuses on the qualitative lift AI provides, such as synthesizing complex insights for decision support or storytelling for the board. Dan cautioned that only measuring time saved risks missing the true, long-term value generated by the second and third buckets, which can transform quality and ultimately impact top-line growth.
To catch up on the entire event, click here.
To learn more about how Pigment is enabling the era of Agentic EPM, click here.
