Simon Woodhead, Founder & Group CEO at Simwood Group PLC

Background

I’ve just read Simon’s recent article, https://simwood.com/2026/01/chapter-3-what-are-you-going-to-build/. Combined with Dave Horton’s post https://blog.jambonz.org/why-were-introducing-a-commercial-license, it’s interesting times in voice AI.  Things are converging as technology matures.

I continue to like the work of Lyle Pratt at Vida on proving out a simple SMB use case, https://blog.tadsummit.com/2024/11/13/lyle-pratt-2/. And fully support Simon’s contention in his article that a) WhatsApp enablement b) voice agents are the answer for most SME (Small Medium Enterprise), and an end of SaaS nickel and diming. We did a TADHack in 2024 on WhatsApp and their bots. https://blog.tadhack.com/2024/11/21/tadhack-2024-sponsors/ that demonstrated this point.

I intro’ed Karel to Dave Horton when Telestax folded, Karel is solving a range of enterprise problems with Voice AI, https://blog.tadsummit.com/2025/12/03/original-gangster-in-voice-ai-karel-bourgois/. The answers are out there, just not well known because of all the BS and crazy fundraising going on.

Here is the session where Simon introduced The Potato last year.

From the beginning in 1996 when Simon founded eSMS / SIMWOOD, the first global SMS gateway between the Internet and mobile phones, a proto-SMS-aggregator. You can read more about Simon’s eSMS adventures in The Definitive Truth in A2P SMS.

Straight into it

Simon frames out how Simwood being 30 years old means its history can be defined by distinct chapters. The first being eSMS where the internet and mobile phones came together; and in that chapter eSMS competed against RIM.

Chapter 2 was the convergence of voice and the internet, with VoIP. Were SIMwood continues to enjoy success. Simon highlights an important distinction between voice and telephony. That’s a distinction the telecoms industry, not users, care about. I use voice across Zoom on this session, have multiple calls over Whatsapp, and occasionally answer a PSTN call even though the PSTN is grossly polluted by robocalling here in the US, that leave bloody voice messages!

Getting into the Meat of Simwood’s Chapter 3

About 3 years ago Simwood began experimenting with AI. ChatGPT had shown AI was now a thing, you could ask AI to create images (e.g. midjourney), and the great experimentation began. From a communications perspective: that included meeting summarization, transcription, sentiment analysis, etc. Which resulted in Simwood’s first voice agent released in April 2025. That made the AI service consumable by a business that used the telephone (telephony) rather than solely a browser.

Simwood is good at learning, it’s how they survived so long. Simon highlighted the crazy valuation of Eleven Labs, and the level of BS in voice AI. So Simwood realized they can make a significant contribution in VoiceAI, based on the fundamentals of owning infrastructure and using open source, which has been their recipe through Chapter 2. I’ve heard similar reflections from other well-respected people across the TADSummit community. It’s not the emperor has no clothes, rather its regular jeans and a t-shirt we all have access to.

It’s a straight-forward formula, there’s a new fashionable technology, that appears quite complicated, but can be made simpler, that by owning the infrastructure can improve the economics, which in turn improves the end user experience. Simwood’s future is in the application of AI, where they will own the solution to compete, differentiate, and build sustainable advantage through listening to and working with the customers across their varied situations.

I like this quote from Simon:

The economics of running GPUs is similar to running TDM gear.

Simon draws an analogy between the current AI boom with the .com boom. While the companies leading the current investment cycle are large established businesses, some of its leadership have a reputation for BS and ignoring people’s concerns. Which means from my perspective, the crash will not be a nuclear winter, selective investment at this point is wise. Would I invest in Eleven Labs today? No, because there are great open source projects and companies that compete directly with less bluster.

AI Agent Development – channel needs

Last week Simwood announced their progressive pivot to AI with a number of improvements in cost and latency. The objective is to be the Vapi for carriers. Vapi AI is a developer-focused platform used to build, test, and deploy fast, conversational voice AI agents, acting as middleware between telephony systems and AI models.

Simon highlights that Vapi wants to serve the enterprises directly, their investors expect this, developers are their channel to enterprises. While carriers want to continue to serve enterprises directly, with enhanced Voice AI services. It reminds me of how Twilio would partner with carriers to offer APIs, e.g. Softbank in Japan. Needless to say, that model was not successful. I remember a statement from the old Twilio CEO on how they do not need sales, they have APIs. They have sales and channels partners to reach enterprises. Reach matters.

Telcos have long established enterprise relationships. Last week in discussing the open source workflow Mation, Sebastian and Marten discussed the importance of the visual tool to aid carrier pre-sales in implementing carrier APIs for enterprise workflows. While organizations like the GSMA and the toadies they surround themselves with can make carriers seem out of touch. Examine the business being done on the ground gives a different perspective. Not marketectures of pointless API standards at conferences. Camara being such an example. Rather witnessing the customer understanding of technical sales.

Simon highlights the role carriers can play in soothing a few rattled nerves after early AI implementation failures. But also the AI agents can remove the awful robocalling plaguing the PSTN in the US and pushing SMB away. Simon shares the story of the guy that installed the TVs in the Simwood office, and the challenges he had in getting telephony working for his business. As Simon frankly explained, even a glorified IVR would be progress. I believe Vida Voice is onto a winner given my experience of being a SMB in the US, its small steps, but using VoiceAI helps small businesses.

AI Agent Development – Ownership

The market always lags technology. Facetime is popular with my wife’s family, though multi-party can result in odd cropping. We use Zoom for my son’s private lessons, I was using Zoom for a virtual consult with my doctor to prep for a procedure. Though Zoom seems to be used less and less in the corporate market, squeezed out by Teams and Google Meet, Zoom is fighting back with a mix including the Zoom phone and specialized corporate packages, see above healthcare example. Zoom’s 2026 revenues should be about $4.85B, so even though some people have moved to Google Meet, it’s a complex, heterogeneous environment.

Simon backs up his point on making it easy / accessible for the carriers’ technical sales to build the agents in minutes. $20 per month for a simple agent, multiplied by the number of SME is a tidy business. I mentioned how the agent can be pointed to the company website to learn the hours of operations, services and pricing, etc. This prompted Simon to discuss what they did in the version 2 of their agent.

• Call control for specific tasks, e.g. when buying a specific service.
• Adding RAG (Retrieval-Augmented Generation) to the agents. We’ve discussed the practicalities of RAG with the Original Gangster in VoiceAI, Karel Bourgois. Which includes live web search. When the customer asks about opening hours, current weather alerts can be included. Which is highly relevant after the past snow storm, Google Maps mentioned in its driving directions. The agent is like chatting with a real person, thanks to the inclusion of live information.

We are in a phase of deploy and see how the customer uses it in their specific situation.

Reality matters, for example with interruption handling. Those experienced with Voice AI are tolerant, customer are not. Latency matters in this situation. Hence ownership of the infrastructure is critical to enable the flywheel of responsive development, which thanks to AI Coding is a small team building amazing services.

This was a fun catch-up, that highlights the convergence happening in VoiceAI, and the importance of ownership and responsive development from working with customers using the agents in their business.