Stillpoint employs only senior experts and hands the rest to agents. There are no junior staff. The productivity this promises still meets a limit set by human beings.

Chris Donahoe, Co-founder of Stillpoint, a corporate affairs firm built to be AI native, argues that the binding constraint on agentic work is now biological rather than technical.

Chris explains that building an AI native firm is an exercise in destroying assumptions, not a technology project, and that many companies use AI merely to speed up an operating model that may no longer deserve to exist. He describes inverting the professional services pyramid: senior human talent only, an entirely agentic support layer, and delegation to teams of agents rather than junior colleagues. The gain has a ceiling, because human reading speed, attention, and judgment have not risen since 2022, so the volume agents produce backs up at the last mile of review. Chris also names a management shift, arguing that even a solo operator now runs a team of agents while no orchestration tool yet exists to coordinate them.

• "It's really an exercise in destroying assumptions": the starting point is questioning why the business works the way it does.

• The talent pyramid inverted: partner and managing director level humans only, no junior staff, an entirely agentic support layer.

• The biological bottleneck: agents take work to ninety per cent done, but customer ready output depends on a person's finite hours and attention.

• "Even solo operators are managers now": scoping, delegation, feedback, and judgment become baseline skills for every professional.

• A candid tools verdict: Claude is strong at formatting Word, Excel, and PowerPoint, yet in Chris's view a poor copywriter, so he routes drafts through ChatGPT and back.

Press play for an account of running a firm where the agents never tire and the people set the pace.

Chapters

00:00 Building an AI native company: destroying assumptions

04:30 Encouraging experimentation with AI tools

07:07 Characteristics of an AI native firm versus traditional

08:13 Talent structure in AI native firms

10:28 Challenges of managing a human-AI hybrid organisation

12:54 The need for orchestration tools in AI workflows

13:54 Skills for managing in an AI native environment

17:03 Management and coordination of AI agents

19:35 Managing interdependencies and orchestration

23:43 Human biological bottleneck and work volume

27:40 Implications of human speed limits in AI workflows

34:04 Recruiting and developing talent for AI native organisations

37:28 AI native firms and future challenges

Ten years in the same sales role now predicts AI resistance. Leo Rogers has seen the pattern often enough to name the mechanism: representatives who hit their number quarter after quarter were never questioned on their method, so the method was never examined. His verdict is that AI adoption is a people problem as much as a technology problem.

Leo Rogers, co-founder and CEO of Curvo, which builds a real-time execution layer for sales teams, argues that knowledge has become a commodity and that the need for generalists is on the rise.

He makes the case for skill stacking, holding roughly 30 per cent of several disciplines, enough to set strategy and to judge an output, with agents supplying the depth. He sets a limit on his own argument: a non-designer who commissions a prototype from an agent still cannot say whether the design system is robust. This means prior exposure remains the gate on quality. He describes orchestration as a generalised agent sitting above specialised agents, deciding which to mobilise and reconciling what they report back, and treats it as the change that makes materially more complex work possible. He also predicts an inversion of the labour market in which agents appoint humans for the last mile, on the grounds that an agent in a sandbox cannot shake a hand or fix a chair.

● The laggard profile: decades of steady growth on the SaaS wave, leadership motivation moving from growth to preservation of the founders’ / managements’ wealth, then private equity arrive with two or three years ‘to squeeze the lemon’.

● Resistance mapped to tenure: one year, three years, five years, and at ten years AI adoption becomes much more challenging, with reps coasting on their black books rather than the CRM.

● Robotics without the humanoid: a bricklaying machine with no legs working at the rate of five bricklayers, set against a humanoid taking half an hour to place a mug in a dishwasher.

● On jobs: Meta and other large technology firms cutting headcount, Leo’s view that organisations carry cost centres agents make redundant, and his warning that universal income becomes a real possibility if AI adaptation fails.

● On executives: checks and balances survive because agents lack empathy and their judgement is geared to be entirely commercial. Middle management, on his account, does not survive.

Leo’s central claim is testable inside a week. Pick one function, ask whether the people in it hold 30 per cent of the disciplines their work touches, and the answer will tell you how ready the organisation is.

Chapters

00:00 AI's Impact on Headcount and Organisational Fat

00:30 Skill Stacking and Cross-Functional Value

00:59 Training Programmes and Upskilling in Large Organisations

02:26 Orchestration of Specialised AI Agents

05:00 AI Adoption in the UK and Change Management

06:10 Resistance to AI Adoption and Entrenched Roles

08:00 Restructuring and Talent Shifts for AI Integration

09:20 AI's Effect on Jobs and Cost Rationalisation

10:46 Future Skills: Broader, Generalist Roles

12:16 Becoming a Human-AI Hybrid: Skills and Exposure

14:23 Critical Thinking and Human Oversight of Agents

16:56 Acquiring Experience in Large Organisations

17:56 Self-Directed Learning and Side Projects

20:12 Characteristics of Lagging Organisations

23:23 Robots and Robotics in Value Creation

25:45 Orchestration of Agentic AI and Marketplaces

28:02 The Human Role in a Future AI-Driven Workplace

30:44 Organisational Transformation and Change Management

Sovereignty, trust and context are often addressed separately, yet there is an underlying interplay between them. Send the same prompt to the same large language model and two institutions receive the same answer, so the advantage that remains is the context a firm owns. That context is arguably the ground on which trust has to be earned, and it holds only if the data and intelligence layer beneath it are not outsourced.

Peeyush Aggarwal, senior partner at Deloitte, makes the case that banks, and organisations across the public, private and third sectors, have to solve all three at once.

He argues that the book of record captures what happened, while the intelligence lives in the why, the how and the who, much of which has never been written down. He shows that agentic AI, software that pursues goals and takes actions rather than answering a single query, breaks the settled architecture of transaction, engagement and analytics, while latency, predictable outcomes and cost remain unsolved. He warns that the real risk in a multi-agent system is not one agent failing but the interaction between them, where intent drifts and no one can trace who decided what. He treats the disintermediation of banks as an open question, with trust the deciding factor.

• "You do not outsource your intelligence layer," a principle Peeyush applies to energy, compute, models and data alike.

• Trust cannot be deployed like software; it is earned transaction by transaction, and a single agent taking a wrong turn or inappropriate action can lose it.

• A customer who says "I lost my wallet" forces four actions at once: detect fraud, issue a new card, block the old one, and offer further protection.

• Four pillars of a durable AI strategy: a choice of tools, a business metric, instrumented trust, and data and context at the heart.

The episode hands leaders a test to apply before approving another agent: name the business metric it is meant to move, and the context it draws on. Press play.

Chapters:

00:00 National sovereignty and AI infrastructure

05:54 Controlling the intelligence layer and infrastructure design

07:51 Data management and the data moat

09:26 Business model evolution with AI

11:37 Impact of AI on value creation in industries

13:17 AI readiness of software architecture

15:41 Challenges of deploying AI at scale

18:39 Consumer trust and AI agents

22:42 Disintermediation of banking and trust

24:06 Ecosystem emergence and innovation

25:50 Key ingredients for a successful AI strategy

29:30 Summary: Trust, context, and strategic pillars

Cost cutting in customer service ended in 30% pay rises. The organisations that put AI agents live in spring 2025 were, by the autumn, paying 30% above market rate to keep the staff their business case had counted as savings. The AI agents took the easy questions and left people who were harder to replace.

Ed Thompson, Senior Vice President for Market Strategy at Salesforce, argues the proof of concept is over and the proof of value is the next frontier.

Around three quarters of companies can evidence time savings while only 10 per cent of finance directors can see a financial benefit, because an employee saving 14 minutes a day goes home on time rather than off the payroll. He calls the figure of an average of 12 agents per organisation true and completely misleading, since one customer in a room, of say 20, may run 1,000 agents live and half run none. He rejects a jobs apocalypse for an uncomfortable reason: many companies lack the capabilities to implement aggressively, as their data is poor and their tech debt, or legacy systems, is worse. And the largest employment effect so far has affected middle managers, through delayering, rather than on the contact centres everyone watches.

• Token prices fell by around 99 per cent over five or six years, then rose for frontier models only.

• Running traditional workflow is 1,000 to 10,000 times cheaper than using tokens for the same work, which splits building from operating.

• Uber burnt through its entire 2026 token budget in four months.

• Z.ai's GLM 5.2 runs at roughly a tenth of the cost per token, and 80 per cent of AI startups use Chinese and open source models.

• In one company, a risk committee of 21 people approved zero agentic tasks in a year.

Ed puts the employment effect a decade away, on the pattern of the loss over time of UK supermarket checkout jobs. The decisions that settle who leads in ten years are taken this quarter, and the token line in next year's budget is one.

Chapters

00:00 The Role of Humans in an AI-Driven World

01:08 AI Adoption Trends in Europe

05:12 The Rise of Agentic Organisations

10:10 Overcoming Adoption Challenges

12:07 Impact on Jobs and Employment

18:13 Understanding Token Economics

22:15 Scaling AI for Value

24:15 The Future of Work and Human Identity

Done well, AI in customer service needs more people, not fewer. That runs against the entire case being made for it. Matt Price has built a company on exactly this claim, and he brings the evidence.

Matt Price, Founder and Chief Executive of Crescendo, argues that the industry has spent twenty years optimising for the wrong thing.

He names deflection, the practice of stopping customers reaching a human, as the metric quietly wrecking service and returns. The vendor paid only for AI resolutions has every reason to keep you inside a bot, and Crescendo instead charges only when the whole journey ends in a satisfied customer. He makes the case that moving care from the back door of a business to the front produces around three times the engagement, and that over half of those extra inquiries come from people who want to buy. He is blunt about the point at which these programmes stall, and it is not the technology.

● Deflection as a twenty-year habit that punishes the customer and misleads the board

● The incentive trap: vendors paid per AI resolution keep you inside the bot

● Three to five points of labour retired for every ten points of automation, on a curve that is not linear

● Lovepop's Trustpilot rating moving from 3.6 to 4.6 in a couple of weeks

● “Where transformation or innovation goes to die is in middle management”

Anyone who has been stuck in a loop begging a bot for a human will recognise the problem. This episode sets out the economics behind it, and the change that fixing the incentives produces.

Chapters

00:00 AI's Impact on Customer Engagement

03:42 The Evolution of AI in Customer Service

06:11 Deflection vs. Engagement in Customer Service

08:40 Aligning Incentives for Successful AI Implementation

11:21 The Labour Dynamics of AI in Customer Service

13:11 The Bolt-On Bot Phenomenon

14:44 Continuous Improvement in AI Deployments

16:59 Real-World Examples of AI Success

21:44 Transformation Practices for AI Adoption

25:15 Key Strategies for AI Implementation in Organisations

Every technologist repeats the same phrase: AI with humans in the loop. André Lacroix inverts it. The future, he argues, is purpose-based, people-centric leadership with AI in the loop, and the difference is not semantic.

André Lacroix, Chief Executive of Intertek Group plc, makes the case that emotional intelligence, not technology, will decide which organisations win the AI decade.

His reasoning starts from an uncomfortable number. Gallup's latest survey shows 80% of the global workforce is not engaged or not actively engaged, which André frames as three billion people arriving at work without purpose. He argues AI's real value is buying leaders back the time to fix that, by automating the analytical work that fills their diaries. He predicts every algorithm and agentic solution will eventually become a commodity, which leaves emotional intelligence as the only durable advantage. And he refuses, flatly, the idea of a CEO avatar answering employees on a leader's behalf.

In this conversation:

• The Gallup figure behind André's claim that the corporate world is "over managed and underled"

• Why he processes three million data points in his role, and calls the goal "data science in real time at scale"

• His four foundations for responsible AI: governance, transparency, security, and functional performance

• His warning that uploading data into a third-party large language model means the data leaves your company

• His argument that instinct is "the result of years of data processing in our brain," and should be trusted, not overridden

André also shares breaking news: a sequel to his book Leadership with Soul, built around the idea he calls EQ supremacy in a world of AI. Press play for a leadership argument that puts AI firmly in second place.

Chapters

00:00 Trusting Instincts and Embracing Change

03:03 Leadership in the Age of AI

10:24 The Role of AI in Decision Making

17:20 Emotional Intelligence vs. AI

22:38 The Future of Work and AI

26:04 Skills for Reinvention in an AI World

27:38 Trust and Transparency in AI Implementation

Large organisations are achieving limited benefits from AI projects. Yet, their appetite to invest more is undiminished. Rob Lunn, Product Manager at Fnality, a regulated wholesale payment system that settles transactions on-chain, argues that this paradox is not just a general business problem. In financial services, where unit cost discipline and regulatory compliance define survival, it is a structural risk that leadership teams have not yet priced in.

Rob makes the case that AI and blockchain are not competing technologies; they serve fundamentally different parts of the payment chain, and confusing the two is one of the more expensive mistakes a payments organisation can make.

The conversation maps the payment process from instruction receipt through to settlement and reconciliation, and locates precisely where each technology delivers value. Blockchain, through atomic settlement, removes counterparty risk at the point where absolute certainty is required: payment occurs if and only if the corresponding asset transfer happens. AI delivers its greatest value at the start and end of that chain, reading unstructured invoice / payment instruction data, extracting meaning from vendor records, and supporting liquidity forecasting and reconciliation. Where existing automation already achieves very high straight-through processing rates, agentic AI faces two barriers Rob identifies as equally significant: the legal framework for autonomous transaction signing, and enterprise and consumer trust.

Highlights from the conversation:

• The paradox Rob names: large organisations report a low success rate with AI projects today, yet investment appetite, based on industry surveys and events including London Tech Week, is off the scale.

• Atomic settlement explained: blockchain links payment and asset transfer so that one cannot occur without the other, removing the settlement risk that became systemic during the Lehman Brothers collapse.

• Settlement risk is one of the factors determining how much capital banks must hold under global regulations, making its reduction a direct financial benefit, not just an operational one.

• Rob's leadership framework: be highly selective in AI use cases; focus on a small number of well-defined projects; assess total cost of ownership against customer outcome, regulatory obligation, and long-term capability building.

• The talent pipeline argument: eliminating junior roles before organisations build the technical and domain knowledge those roles develop leaves the future management layer without the skills to run AI-dependent operations safely.

For leaders in financial services asking where AI and blockchain actually earn their place, and what the cost of getting that wrong looks like, this conversation provides an unusually precise answer.

Chapters

00:00 Fnality and Blockchain in Payments

03:23 Blockchain's Role in Mitigating Settlement Risk

08:31 Synergies Between AI and Blockchain Technologies

11:07 Limitations of AI in Payment Processes

14:52 Regulatory Challenges in Financial Technologies

17:23 Impact of AI and Blockchain on Leadership in Payments

21:12 Long-term Challenges in the Payments Industry

23:21 Preparing for AI and Blockchain Integration

Taiwan makes 95% of the world's AI servers. The rest of the world runs on what Taiwan builds.

Sega Cheng, Co-Founder and Chairman of iKala, argues that Taiwan's hardware leadership is the clearest illustration of how foundational AI infrastructure has become: one country manufactures, assembles, and supplies the chips and servers that power AI for virtually the entire world. From that vantage point, he makes the case that intelligence is becoming the third utility of human civilisation, after water and power. Just as no enterprise or state would accept indefinite dependency on another for electricity, he argues that home-grown AI capability, at the level of data, models, and infrastructure, is no longer a strategic preference. It is a necessity.

That argument is the starting point for a conversation that moves well beyond Taiwan. The tension Sega identifies is sharp: the country building the infrastructure for global AI has yet to fully use it itself, with more than 70% of Taiwanese businesses still to genuinely integrate AI into their operations. Organisations everywhere are making the same structural mistake: layering AI on top of legacy systems and legacy mindsets, then wondering why the returns do not materialise. His answer is not transformation at scale. It is something more disciplined: start where the results are visible. In iKala's experience, that means marketing, where a two-point improvement in targeting or a 20% lift in social media performance gives an organisation concrete evidence of AI's value before committing to deeper and more disruptive change.

The conversation also covers an argument Sega makes with considerable force: what he calls 'consider software soft.' The cost of producing software is approaching zero as AI coding tools advance, which shifts value away from code itself and towards sector-specific application. The organisations that will benefit are those with deep domain knowledge in fields such as agriculture, medicine, and manufacturing, ready to apply rapidly produced software to problems they already understand better than anyone else.

Highlights from this conversation:

• Taiwan assembles over 95% of global AI servers, making it the backbone of the world's AI infrastructure

• 80% of AI adoption effort goes into data collection and cleansing; only 20% touches algorithms or machine learning

• China's AI adoption is state-led and follows a different pattern from the 30/70 split visible in the rest of Asia

• Edge AI is moving from concept to deployment in manufacturing, warehousing, and defence

• iKala's Kolr platform tracks data on over 300 million influencers, providing a concrete case study in AI-layered transformation

Chapters

00:00 AI as the Third Utility

02:32 iKala and AI Adoption

05:10 AI Adoption in Taiwan

08:21 Industry-Specific AI Adoption

09:32 Regional AI Adoption Trends

12:14 Challenges in AI Integration

16:08 Measuring AI ROI

21:16 The Future of Software in AI

24:15 Understanding Edge AI

29:39 Taiwan's Semiconductor Advantage

33:28 Sovereignty in AI and Chip Manufacturing

AI tools at Bank of America are already delivering around 20% productivity improvements in coding. That figure is live, not projected and from the executive responsible for technology and operations in the bank's French entity. He argues it will increase as the technology matures. Yet his most insistent point is this: at every consequential decision point, a human must remain in the loop. The productivity gain and the human override are not in tension; they are the same strategy.

Andy Price, Head of Technology and Operations at Bank of America's French legal entity, makes the case that the difference between organisations that merely adopt AI and those that lead with it comes down to culture, governance, and disciplined application of standard business case logic, not the sophistication of the tools themselves.

The conversation covers Bank of America's seven-year AI journey, from the launch of Erica in 2018 through to its current portfolio of internal and client-facing tools. Andy explains the bank's prioritisation framework: high-volume, repetitive, manual processes come first, assessed against the same return-on-investment criteria as any other technology investment. He argues that AI deployed without stringent governance will go off track, and that human oversight is non-negotiable, not as a regulatory concession, but as an operational principle. In an environment where AI is accelerating into research, drafting, code generation, and client interaction, that principle has direct implications for how far AI autonomy can extend into consequential execution.

Highlights from the conversation:

• Bank of America's AI-assisted coding tools are producing productivity gains of around 20% already, with Andy Price expecting that figure to rise as the technology develops.

• Erica for Employees has measurably reduced technical support calls and response times across the organisation, with automatic escalation to a human when the AI cannot resolve an issue.

• AskGPS searches thousands of internal documents in real time, giving employees faster, more accurate answers to client queries without leaving the workflow.

• Andy's boardroom advice centres on business case discipline: do not automate a process that saves 30 minutes a year and costs significant investment to build. Focus on the repeatable, the high-volume, the genuinely time-consuming.

• On the limits of AI autonomy: "There must be a human being who has to look at it, review it, press the button."

For any organisation asking where AI is actually delivering, and how to build the governance structures that make that delivery sustainable, this conversation provides concrete answers from one of the world's largest financial institutions.

Chapters

00:00 Embedding AI into Culture and Operations

02:10 The Importance of AI Investment in Banking

05:06 AI Outcomes for Clients and Employees

07:43 Leading vs. Simply Adopting AI

09:00 Prioritising AI Investments

10:05 Deployment Strategies for AI Technology

11:10 Reshaping Employee Experience with AI

13:13 Productivity Improvements through AI

14:39 Talent Development and Upskilling

16:09 Responsible and Ethical AI Practices

18:21 The Future of AI in Banking

20:38 Advice for Companies Starting Their AI Journey

Hallucination in AI is not always a failure to be fixed. For some organisations it is a design choice. That observation, made on stage at Brunel University of London, is one of nine tensions this conversation surfaces, examines, and poses as challenges to be resolved as we live and work with AI.

This is the Season 3 opening episode of The AI Adoption Podcast, recorded live in front of an audience at Brunel's Research Festival 2026. Five panellists joined me to discuss The Future With AI: Zahra Bahrololoumi CBE, President and CEO of Salesforce UK and Ireland; Lord Tim Clement-Jones, Liberal Democrat spokesperson for Science, Innovation and Technology and co-chair of the APPG on AI; Janusz Marecki, CEO and co-founder of FractalBrain; Maggie Sarfo, founder of Meres Consulting; and Alex Dalman, Managing Partner at Faith.

The theme is the future WITH AI. Not the future of AI. The emphasis matters. The panel assumes AI is already ubiquitous and asks what life looks like inside that reality, and where the tensions lie that policymakers, business leaders, and academics need to face up to.

Five tensions to listen for:

• Salesforce's agentic AI now resolves between 82 and 84 per cent of complex customer queries on its help platform without human intervention, saving $100 million a year with no engineers displaced. Alex Dalman observes that most organisations have no real understanding of the effort required to get there.

• Janusz Marecki argues that current large language models are unreliable by design: the underlying architecture samples from a probability distribution rather than reasoning from verified knowledge, and the training data itself contains what he describes as inaccurate, harmful, and toxic content. Zahra Bahrololoumi argues that people are smarter than we give them credit for, and that enterprise AI built on curated, governed data is a categorically different proposition.

• 61 per cent of people in business are already using AI without having received any formal training. The government has committed to training 10 million people by 2030. The gap between those two assertions is where a great deal of the risk currently lives.

• Lord Tim Clement-Jones describes graduates encountering only automated assessments throughout the early stages of job applications, with no human contact and no feedback until they clear several initial hurdles. Maggie Sarfo argues that AI should be making us more human by releasing capacity for emotional intelligence. Both observations are accurate about different parts of the same reality.

• AI, as Alex Dalman notes, is unlike every previous technology revolution in that it has no tangible form most people associate with it directly. The mobile phone was something you could hold. AI's experiential invisibility is both the source of public distrust and the condition that allows it to embed everywhere without friction.

There are four more tensions in this conversation. Listen to find them.

Chapters

00:00 The Future With AI: Opportunities and Risks

08:43 SMEs and AI Adoption

12:49 Creativity in the Era of AI

17:07 Societal Readiness for AI: Trust and Responsibility

30:59 Business and Societal Readiness: A Unified Approach

32:17 Assessing AI Maturity

36:17 The Rise of the Agentic Organisation

44:02 Accountability in AI Deployment

55:11 The Impact of AI on Jobs and Skills