Building Trustworthy AI: William Tunstall-Pedoe’s Vision for Enterprise Adoption

The Trust Deficit in Enterprise AI

William Tunstall-Pedoe, the mastermind behind Amazon’s Alexa, is tackling what he identifies as the fundamental barrier to AI adoption in business: trust. Through his new venture Unlikely AI, he’s building a platform that addresses the reliability concerns preventing widespread enterprise implementation of artificial intelligence. “Trust is a theme that is coming up all the time in the industry with artificial intelligence and the major reason why AI isn’t being adopted by businesses,” Tunstall-Pedoe explains during our conversation at a London café.

The High Stakes of AI Implementation

While businesses face increasing pressure from leadership to adopt large language models (LLMs), the reality of implementation has been disappointing. Tunstall-Pedoe cites an MIT report indicating that 95% of generative AI pilot projects at companies fail. The core issue? Current AI systems’ tendency to produce inconsistent or inaccurate outputs.

“If you’re trusting a business process to this tech and it goes wrong one time in 30, 50, 200, that can be very expensive and costly to your brand, your finances and put you in breach of regulation,” he emphasizes. This reliability gap becomes particularly problematic when AI systems handle thousands of decisions daily, where even a small failure rate translates to numerous errors., as additional insights, according to recent innovations

Unlikely AI’s Three-Pillar Solution

Unlikely AI addresses the trust issue through three core principles designed to make AI outputs reliable enough for business-critical applications:, according to market analysis

• Accuracy: Every output must be correct and verifiable
• Explanation: The system must provide thorough reasoning for how it reached its conclusion
• Consistency: Identical queries must produce identical results every time

This last principle is particularly crucial. “There’s nothing more trust damaging than seeing this AI system that you’ve built giving two different answers for the same data when you run it twice,” Tunstall-Pedoe notes.

Real-World Applications and Early Adoption

Unlikely AI is already testing its technology with significant partners. The company has announced collaborations with Lloyds Bank for customer support applications and insurance group SBS for scaling insurance claims processing. Another promising application automates accounting disclosures, a traditionally “massively tedious” task performed by junior accountants.

The platform’s interface resembles familiar chatbots, but with crucial differences in reliability. In demonstration scenarios, the system not only provides accurate answers but also explains the reasoning behind its conclusions and references specific policy sections or regulations that support its response.

Targeting High-Stakes Industries

Unlikely AI is initially focusing on sectors where the cost of error is particularly high. “The low hanging fruit for what we’re selling are high stakes industries where the cost of something going wrong is very high,” Tunstall-Pedoe explains. The company plans to implement an outcomes-based pricing model, aligning its success with customer success.

While currently well-funded after a $20 million raise in 2022 from investors including Amadeus Capital Partners and Octopus Ventures, the company anticipates another funding round next year to support scaling efforts.

The European AI Ambition

Tunstall-Pedoe remains committed to building a significant European technology company. “We want to be a very big, standalone business,” he states, acknowledging the challenge of reaching trillion-dollar valuation in a region that currently has no companies at that scale. Europe’s most valuable company, ASML, stands at approximately €342 billion, while the US hosts nine trillion-dollar companies.

He sees the UK as an ideal foundation for this ambition, citing the talent pool, world-class universities, and business-friendly environment. Regarding recent massive US tech investments in UK AI infrastructure, he comments: “I think the UK has got the potential to be an AI powerhouse… These big tech companies will benefit substantially from that investment.”

Regulation and Responsible AI Development

On the topic of government regulation, Tunstall-Pedoe recognizes the delicate balance required. “There’s a tension around safety and not having regulation get in the way,” he observes. “I think the government genuinely wants the UK to be an AI superpower, so is genuinely concerned about getting it wrong and having that prevent growth.”

He advocates for regulation that “accelerates trustworthy AI adoption,” arguing that properly addressing trust concerns will naturally speed up enterprise adoption across industries.

Building a Legacy Beyond the Product

Beyond creating a successful company, Tunstall-Pedoe is fostering an environment that encourages entrepreneurship among his team. “One of the things I’ve discovered about my startup is a large percentage of the staff are interested in startups, have long term career ambitions to launch their own companies,” he notes.

This approach mirrors successful tech companies like OpenAI and DeepMind, where alumni frequently launch their own ventures. For Tunstall-Pedoe, who knew he had “one or two big startups left” after Amazon, creating this ecosystem represents another dimension of success.

The Future of Enterprise AI

As businesses continue to navigate AI adoption, the trust issue remains the central challenge. Unlikely AI’s approach—focusing on reliability, explainability, and consistency—addresses the fundamental concerns holding back widespread implementation. While the platform is still in early stages, its focus on high-stakes applications and measurable outcomes positions it uniquely in the crowded AI landscape.

The success of ventures like Unlikely AI could determine how quickly artificial intelligence transforms from a promising technology into a reliable business tool across global enterprises. As Tunstall-Pedoe concludes, “if you get the trust bit right you accelerate adoption.”

This article aggregates information from publicly available sources. All trademarks and copyrights belong to their respective owners.

Note: Featured image is for illustrative purposes only and does not represent any specific product, service, or entity mentioned in this article.