I sat down with an AI founder who believes companies are spending first… and asking what it cost later.

Lemuel Low is building Ren3 to help businesses understand, control and optimise the rapidly growing cost of AI. As enterprises deploy copilots, autonomous agents and increasingly complex workflows, many are discovering that traditional software budgets simply do not work when every outcome carries a different and potentially uncapped token cost.We spoke about why AI never became as cheap as promised, who should be accountable when an agent fails, whether companies can predict the cost of a workflow before running it, and why governance needs to be built into infrastructure rather than left as another policy document.

Lemuel also explains model routing, open-weight alternatives, outcome-based pricing, the conflict between model providers and their customers, and what cost-efficient companies are already doing differently.

This is one of the most practical conversations I’ve had about the economics behind enterprise AI. If you care about AI agents, enterprise software, token costs, governance, model routing, pricing, or building sustainable AI businesses, watch this one till the end!

IN THIS EPISODE

00:00 | Teaser

00:51 | Uber's AI Budget & Token Spend

03:32 | Why Token Maximisation Suddenly Stopped Working

05:33 | When Finance and Engineering Collide Over AI Costs

09:42 | The Two Things Every Company Must Control

10:59 | Who Takes Accountability When Agents Fail?

12:40 | Accountability Beyond Governance Frameworks

13:21 | Why the Best AI Model Is Often the Wrong Choice

16:46 | Can You Price an AI Task Before Running It?

20:40 | Inside Reentry’s AI Cost-Control Infrastructure

22:26 | Can AI Companies Really Charge for Outcomes?

26:09 | What Cost-Efficient Companies Look Like

34:24 | Will AI Competition Actually Make Models Cheaper?

35:41 | The Coming Battle Over Enterprise AI Spending

Lemuel Low

Lemuel Low is building Ren3, an AI platform focused on helping businesses manage, attribute and optimise the cost of running AI workloads. His work explores how companies can move beyond unrestricted token consumption by introducing model routing, workload planning, traceability and governance directly into their AI infrastructure.

Through Ren3, he is working to give organisations greater visibility into how tokens are being consumed, which teams and processes are responsible for that spending, and whether the outcomes being produced justify their cost. He believes that as AI adoption accelerates, companies will need to treat token economics as a core operational discipline rather than relying on model providers to control spending for them.

Rahul ThayyalamkandyDirector of Comms & Community, Vertex Ventures SEA & India

Rahul Thayyalamkandy joined the SEA/India team as the Director of Communication & Community in 2024 and is based in Singapore.

Before joining Vertex, Rahul served as a Product Manager at BLOCK71, an incubator and joint venture of NUS Enterprise and Singtel Innov8, where he developed scalable services for portfolio startups and designed mentorship programs. He also gained experience as a Summer Associate at Cocoon Capital, working with the Managing Partner on investment deals.

https://www.linkedin.com/in/rahulthayyalamkandy/

—

Vertex Ventures Southeast Asia & India is a pioneer in investing in tech start-ups in this region and has helped to build a number of unicorns. In this podcast, we seek to share and uncover the Hard Truth — raw, unfiltered insights and venture capital knowledge from across Southeast Asia and India. Tune in to learn from leading founders, innovators, venture capitalists, and industry experts in the region and gain industry insights from those in the know.

If you like what you hear, please follow or subscribe to the show so you don’t miss out on any of our upcoming discussions.

For more information, please visit: https://www.vertexventures.sg