2026: Philosophy, Psychology and the Year Ahead for AI

TL;DR

Back at the desk after the break and I’m shifting focus a little in 2026. Expect fewer technical posts and more educational, thought-provoking content that weaves in bit of philosophy and psychology alongside practical strategy. Looking back, 2025 was transformative for AI adoption - but also revealed serious implementation problems, with some sources indicating 42% of companies abandoning their AI initiatives. This year, we’ll help clients navigate both the opportunities and the growing need to refactor what was built in haste.

Wow, how time flies. Back at the desk after Christmas and New Year celebrations and it’s already late January before I’ve penned (okay, typed) my first blog post of the year.

I wanted to kick things off with a slight shift of focus for 2026 - at least for my posts. I don’t know about you, but my LinkedIn feed is littered with too much technical “look at me” content for my tastes, and posts written purely for SEO/GEO reasons. I’ll try to keep posts educational, insightful, entertaining and thought-provoking this year. Well, that’s the intention anyway.

What Actually Changed in 2025?

Before looking ahead, it’s worth acknowledging what genuinely changed last year. The AI story felt like it was less about new capabilities - it was more about adoption curves finally catching up with the technology.

Businesses stopped asking “should we use AI?” and started asking “how do we implement this properly?” That’s a meaningful shift. Enterprise AI adoption reached 78% of organisations using AI in at least one business function, up from 55% in 2023. And frankly, that still feels like an underestimate.

We saw Claude, GPT-4 and other models mature into platforms you would genuinely want as part of your operational infrastructure. The agentic workflow conversation moved from theoretical to practical - we delivered several projects last year where AI handles complex multi-step processes that would have required dedicated staff just 18 months ago.

But we also saw the first serious signs of what I’d call “AI debt” - and the statistics here are sobering.

What Is AI Debt?

According to S&P Global’s 2025 survey, 42% of companies abandoned most of their AI initiatives this year - a dramatic spike from just 17% in 2024. The average organisation scrapped 46% of AI proof-of-concepts before they reached production. I’m all for a bit of fail fast, learn and evolve, but that’s a big chunk of projects!

MIT’s research puts it even more starkly: despite $30-40 billion in enterprise investment, only 5% of AI initiatives are producing measurable returns. Ouch! The pattern is consistent - companies attempted to force generative AI into existing processes with minimal adaptation, and 95% of projects failed to demonstrate profit-and-loss impact.

The gap between “working demo” and “production-ready system” is wider than many vendors acknowledge.

Two Examples of AI Debt in Practice

The customer service chatbot problem: Many businesses rushed to deploy AI chatbots in 2024, only to discover the systems couldn’t handle real-world complexity. 75% of customers feel chatbots struggle with complex issues and fail to provide accurate answers. Resolution rates vary wildly - from just 17% for billing issues to 58% for straightforward returns. Mind you, plenty of human operators out there seem to struggle with this too.

The infamous cases made headlines: DPD’s chatbot apparently started swearing at customers and writing poems criticising the company after a system update. A Chevrolet dealership’s bot agreed to sell a 2024 Tahoe for $1 and claimed it was “legally binding - no takesies backsies.” These are extreme examples, but they illustrate what happens when AI systems are deployed without adequate governance, testing, or human oversight.

The sales automation graveyard: MIT found that more than half of generative AI budgets went to sales and marketing tools - yet ROI was lowest in precisely these areas. The biggest returns came from back-office automation: eliminating business process outsourcing, cutting external agency costs, and streamlining operations. Many organisations are now discovering that their shiny sales AI tools generate impressive-looking activity metrics whilst failing to move the needle on actual revenue.

The common thread? Implementations that prioritised speed over sustainability, impressive demos over practical workflows, and vendor promises over measured outcomes.

What Lies Ahead in 2026?

My prediction for this year isn’t that AI gets better (that’s obvious) - it’s that we enter a consolidation phase.

Expect fewer tools doing more, workflow orchestration becoming the real differentiator, and growing demand for people who can refactor the messy implementations of 2024-2025 into something maintainable. Gartner predicts that over 40% of agentic AI projects will be cancelled by the end of 2027 due to escalating costs, unclear business value, or inadequate risk controls.

For Aandai, that plays to our strengths. We’ve always prioritised practical, maintainable solutions over impressive-looking demos that fall apart in production. The organisations that succeed with AI share common characteristics: they redesign workflows before selecting tools, they invest heavily in data quality, and they measure actual business outcomes rather than activity metrics.

A Philosophical Counterpoint

But let’s mix it up a bit. I’m going to interleave posts about technical delivery, strategy and the like with a few more philosophically and psychologically focused articles moving forward.

Having studied philosophy at King’s College (a fair few years ago now), I’ve always found it a mentally nourishing counterpoint to technical development and marketing strategy. Hopefully I can share some thought-provoking overlaps between that and our world of work today.

Daniel Kahneman’s Thinking, Fast and Slow remains foundational for anyone in marketing - the tension between System 1 (fast, intuitive, emotional) and System 2 (slow, deliberate, logical) explains why purely rational marketing so often fails. We like to think our customers weigh evidence carefully before making decisions. They don’t. Neither do we. If you haven’t read it, give it a go. Not exactly easy reading, but a fascinating exposé on cognitive biases, and some entertaining head-scratchers such as the bat-and-ball question.

Rory Sutherland extends Kahneman’s ideas brilliantly into marketing practice. His book Alchemy argues that the most effective solutions are often psychologically rather than logically optimal - a useful corrective when you’re drowning in data and dashboards. His central insight: we spend too much time trying to change what people do, and not enough time understanding why they do it.

Richard Thaler’s concept of “choice architecture” - how the structure of decisions influences outcomes - applies to everything from landing page design to how you present options in a proposal. His book Nudge (co-authored with Cass Sunstein) offers practical frameworks for designing environments that help people make better decisions without restricting their freedom.

Nassim Taleb’s “antifragility” offers a useful lens for building automation that improves under stress rather than breaking unpredictably. His work in Antifragile helps frame the difference between systems that merely survive disruption and those that actually benefit from it - directly applicable to how we design workflows that handle edge cases gracefully.

And it’s likely that my personal passion for some of the great existentialist philosophers may creep in at points. Bear with me, there’s a connection. Merleau-Ponty’s work on embodied cognition has surprising relevance to how we design human-AI collaboration - how tools become extensions of ourselves rather than obstacles to work around. Simone de Beauvoir’s ethics of situated action offers a framework for the “should we automate this?” questions that increasingly arise in client conversations. Not every efficiency gain is worth pursuing.

Why Psychology Matters for Marketing

And psychology… well, that underpins every successful marketing campaign, creative delivery and frankly, every single business meeting we have. I’m not suggesting we all become NLP adepts, but it pays dividends to take a step back and refresh your grasp of basic principles every once in a while.

These aren’t abstract academic interests. They’re practical frameworks that inform how we approach client problems.

Understanding that people make decisions emotionally and justify them rationally explains why feature-focused marketing underperforms benefit-focused messaging. Recognising that loss aversion is roughly twice as powerful as equivalent gains shapes how we frame proposals. Knowing that people anchor on the first number they see influences how we present pricing options.

The best marketers I’ve encountered aren’t necessarily the most technically skilled - they’re the ones who understand how people actually think, decide and act. The psychology comes first; the tactics follow.

What Lies Ahead for Aandai?

Well, it looks set to be a year of solid growth - some really exciting projects already signed up and scheduled in - so expect a somewhat patchy blog writing schedule as things get busy.

But I’m hopeful we’ll broaden the team as the year progresses, so I’ll always find the time to stay on top of the latest developments relevant to your business needs. We’re increasingly focused on helping organisations navigate the AI implementation landscape thoughtfully - building systems that actually work in production, not just in demos.

Looking forward to an even brighter and better year ahead.

FAQs

What is AI debt and why does it matter?

AI debt describes the accumulated technical and operational problems from hastily-implemented AI systems. Like technical debt in software development, it compounds over time. With 42% of companies abandoning AI initiatives in 2025 (up from 17% in 2024), many organisations are discovering their early implementations need serious refactoring.

Why do most AI projects fail?

According to MIT research, 95% of generative AI pilots fail to deliver measurable P&L impact. Common causes include poor alignment between technology and business workflows, inadequate data quality, and attempts to force AI into existing processes without adaptation. Projects that succeed tend to focus on back-office automation rather than sales and marketing.

How does behavioural economics apply to marketing?

Behavioural economics reveals why purely rational marketing often fails. Daniel Kahneman’s distinction between System 1 (fast, intuitive) and System 2 (slow, deliberate) thinking explains how customers actually make decisions. Rory Sutherland extends this into practice, arguing that psychologically optimal solutions often outperform logically optimal ones.

What changed in AI adoption in 2025?

2025 marked the shift from AI as novelty to AI as infrastructure. Enterprise adoption reached 78% of organisations using AI in at least one business function. However, failure rates also increased - 42% of companies abandoned most AI initiatives, up from 17% in 2024, revealing the gap between implementation hype and operational reality.

What AI trends should businesses watch in 2026?

Expect consolidation rather than proliferation - fewer tools doing more, workflow orchestration becoming the differentiator, and growing demand for refactoring poorly-implemented 2024-2025 systems. Gartner predicts over 40% of agentic AI projects will be cancelled by 2027 due to escalating costs, unclear business value, or inadequate risk controls.

About the Author

Marc Price is the founder of Aandai, a B2B automation and AI consultancy helping mid-market businesses streamline their go-to-market processes. With 24+ years in B2B technology and a physics and philosophy degree from King’s College London, Marc brings both technical expertise and a broader perspective to automation strategy. He’s a firm believer that the best technology implementations are grounded in understanding how people actually think and work.

Want to Discuss Your AI Strategy?

If you’re wrestling with AI implementation decisions - or suspect your existing automations might benefit from a fresh perspective - we’re happy to talk it through.

Book a free 20-minute consultation to discuss your specific situation.