AI is Everywhere but Value is Not | Passion Labs' Annual AI Review 2025

Blog

Salome Tirado

Marketing Lead

AI Adoption: Experimentation Up, Impact Down

‍

We’ve consolidated the latest research from McKinsey, Forrester, OpenRouter and Forbes, alongside what we see inside the organisations we work with every day, and this is our summary. 2026 is not about buying more AI tools, it’s about understanding the systems you’re building. The pattern is consistent: AI is everywhere but value is lagging. Most companies are experimenting, yet few are scaling in a way that truly transforms the business. This report brings together industry research and real-world experience to help leaders find clarity amid the noise.

‍

‍(Scroll down to access the free 2026 AI Strategy Guide) .

‍

‍

What is Happening?

‍

McKinsey’s latest global survey found that 88% of organisations now use AI in at least one business function, up from 78% last year. Yet almost two-thirds have not moved beyond pilots or proofs of concept. In other words, the tool might work, the demo might impress but the organisation doesn’t change. Only about one-third of firms have started scaling AI across the enterprise and even among “successful” adopters, just 39% report any measurable EBIT impact from AI at the enterprise level. In plain terms, over half of companies investing in AI aren’t seeing real money returns from it. Most “AI wins” remain isolated use-case level achievements, not broad transformation.

‍

Why is this happening?

‍

Often, companies chase the latest AI tools without a clear problem to solve (a pattern we also saw in the dot-com era’s over-investment). There’s talk of an “AI bubble,” but the danger isn’t the technology itself, it’s the “random acts of AI” with no strategic system behind them. The lesson is becoming clear: AI adoption alone doesn’t guarantee value. To capture ROI, you need to rewire how your organisation works.

‍

The Workflow Redesign Imperative

‍

If there’s one thing that separates AI leaders from the rest, it’s this: high performers focus on workflows, not just model outputs. McKinsey identified a small cohort (~6%) of “AI high performers” – companies getting 5%+ EBIT uplift from AI. What do they do differently? For one, they are ~3× more likely to fundamentally redesign workflows around AI, rather than simply bolt AI onto existing processes. In these successful initiatives, teams don’t just ask “What can this model do?” – they ask “How should our process change, given what AI can do?” It’s a crucial difference.

‍

This is why buying more AI tools alone rarely boosts productivity: you cannot patch a broken process with a smart tool. Real gains come when you rethink the process itself to fully leverage AI. Not coincidentally, the highest performers also pursue AI for growth and innovation (new products, services) – not just cost-cutting. They treat AI as a business transformation lever, not a shiny demo.

‍

AI is only as transformative as the process you apply it to.

‍

The takeaway: before deploying AI at scale, map out your workflows end-to-end. Identify where decisions bog down, where data isn’t flowing, where manual effort adds no value. Fix that design first. Our experience echoes the data – every “failed” pilot we’ve seen wasn’t due to bad AI tech; it was due to a mismatch with organisational reality. The tool couldn’t fix the system because the system had to change first. This is why custom end-to-end, like the ones Passion Labs’ implements often outperform off-the-shelf products. It ensures that the solution fits the organisation’s real context (integrated data, user experience, business logic).

‍

AI Agents: Rising Fast but Not Plug-and-Play

‍

What is happening?

‍

One of the biggest shifts in 2025 was the emergence of so-called “agentic AI” – AI systems that can reason, take multi-step actions, or call tools autonomously.

‍

‍

OpenRouter’s analysis of 100 trillion tokens of real-world AI usage points to a clear behavioural shift. By late 2025, over 50% of all AI queries were routed to “reasoning” models (which can plan and iterate, not just answer). Teams stopped using AI just to generate text and started using it to execute tasks and logic. Similarly, requests that invoked tools or external actions grew to ~15% of usage, indicating more AI being used in connected workflows rather than standalone Q&A. In essence, people are trying to use AI to work, not just to write.

‍

However, there’s a reality check: while everyone talks about “AI agents” taking over workflows, very few organisations are ready to deploy true autonomous agents at scale.

‍

'About 60% of firms are experimenting with AI agents, but fewer than 1 in 10 have scaled an agent use-case in any function'- McKinsey 2025

‍

Agents only create value when they deeply understand your world – your data, systems and rules. That means successful agents will likely be custom-built or heavily tailored to each organisation. If you simply buy a generic “AI assistant” and expect it to run your internal processes, you’ll be disappointed. As our CSO Dr Nadine Kroher puts it, companies should customise models with smarter techniques e.g. using retrieval-augmented generation (RAG), in-context learning, or multi-agent frameworks. These approaches are far more efficient, flexible and often just as powerful.

‍

You can watch our full webinar here where she broke down the differences between AI SaaS and Custom Builds starting from the basics.

‍

This is a big opportunity for those willing to invest: narrow, well-designed agents (think of an AI that handles your customer onboarding, or your inventory planning) can deliver huge efficiency gains. But it requires connecting the AI to your proprietary data and providing clear guardrails. In 2025, we specialised in a range of custom projects for clients. (You can view examples of these projects here) .

‍

The bottom line: agentic AI is coming but integration and customisation are the name of the game. It’s better to start structuring your data and processes for AI now, rather than waiting for a magic off-the-shelf agent that may never fully understand your business.

‍

The Human Factor: AI Literacy, Trust and “AIQ”

‍

All the research agrees: AI doesn’t replace people – it replaces tasks. Yet many employees remain uneasy or unprepared to use AI. Forrester reported a stunning statistic: 30–60% of employees with access to an AI tool never use it, often because they don’t understand it, don’t trust it, or fear making mistakes. That’s a huge waste of potential. The analysts warn that underinvesting in training and change management will undermine your entire AI strategy. They introduced the concept of “AIQ” (AI Intelligence Quotient) – essentially, your workforce’s ability to understand AI’s capabilities, use cases, ethics and limitations. If your organisation has low AIQ, even the best tech will flop. Employees may either misuse the tools or avoid them altogether, perceiving AI as a threat or a confusion rather than a benefit.

‍

‍

The importance of AI literacy

‍

Building AI literacy and trust spans formal training (e.g. teaching what generative AI can and can’t do, how to craft good prompts, how to interpret AI outputs) and cultural work (addressing fears, celebrating AI wins, creating a safe environment to experiment).

‍

Forrester’s data shows that when employees are properly supported, adoption jumps dramatically (in one study, from 25% to 76% usage when training was provided).

‍

The importance of AI inclusion

‍

Involve employees in designing AI workflows so they feel ownership. As Forbes and others have noted, companies that help their talent grow in an AI-first environment will attract and retain the best people – because those people see a future for themselves in the organisation. Conversely, if workers feel AI is something being “done to them” in secret, resistance will grow. The key is to frame AI as an “opportunity builder” for employees (to use Forrester’s phrase). E.g. Show how it can automate the drudge work and enhance their roles (not eliminate them).

‍

When people see AI helping them shine – making better decisions, focusing on creative/high-value work – they become your biggest champions.

‍

We’ve seen this first-hand: in one client workshop, a skeptical team went from fear to excitement after a half-day of hands-on AI demos with their own data. Empower your people, and they’ll amplify your AI investment rather than undermining it.

‍

How should you respond to these insights? Here are six strategic priorities we recommend for 2026, based on the data and our on-the-ground experience:

‍

‍Start with systems, not tools. Don’t ask “What else can ChatGPT do for us?” – ask “How does work get done here and where could AI improve the flow?” Map critical workflows end-to-end and reimagine them with AI in the loop. A mediocre tool in a well-designed system beats a great tool slapped onto a broken process. Focus on judgment points (for human-AI collaboration) and automation points (for straight-through processing).

‍

Think custom before generic. For quick wins, a generic AI app (writing assistants, chatbots, etc.) can help in small ways. But the breakthrough productivity often comes from custom AI solutions fitted to your unique needs – whether it’s a tailored model on your proprietary data or a bespoke integration into your software.

‍

Off-the-shelf AI has broad but shallow impact; custom AI can dig deep into your value chain and create defensible advantages.

‍

Invest in your people’s “AIQ.” Make AI literacy, training and experimentation a core part of your culture. Educate teams on how AI works, its strengths/limits, and responsible use. Encourage sandboxes and pilot projects where employees can get hands-on. The goal is to turn fear into curiosity and confusion into confidence. When employees trust the AI tools (and know how to use them), adoption will no longer be a hurdle – it will be a pull from the business, not a push from the top.

‍

Build for agents (but get your house in order). AI agents (software that can autonomously perform tasks) will mature in 2026. Prepare by cleaning up data silos, documenting processes and implementing strong governance. Start with “co-pilots” that assist humans in multi-step tasks; gather learnings on where autonomy truly adds value. Be realistic: an agent that can reset a user’s password or triage an email queue is far more achievable (and useful) than a mythical AI CEO. Plan for incremental autonomy.

‍

Leverage the bubble to your advantage. We are in a feverish period of AI investment and hype. Use this to incentivise action in your organisation. Budgets are available and stakeholders are interested. However, stay focused on your strategy (avoid adopting tech for hype’s sake). This is an ideal time to build foundational AI capabilities (data pipelines, MLOps, model libraries) and experiment with new ideas. The key is to capture value from each experiment (even if just learning), so you’re not left with a graveyard of pilots.

‍

Design for change. The AI landscape will likely shift in the next 12 months – new models, regulations, maybe new paradigms (we’re watching concept-based models and multimodal systems on the horizon). Architect your AI solutions with modularity. For example, use an abstraction layer so you can swap out the NLP model underneath, or design your workflow so that if a tool API changes, your whole process doesn’t crumble. In short, build adaptable systems. This ensures you benefit from innovation rather than being broken by it.

‍

If this feels like a lot to take in, that’s understandable. At Passion Labs, our role is to help organisations do exactly this end to end. We design and deploy AI systems that deliver measurable impact, combining deep research with practical execution. As an R&D lab of PhD-level scientists and engineers, we build real-world, domain-specific solutions. We’re able to look at your business holistically, benchmark it against the state of the market and move from strategy to deployment with confidence. If you want your 2026 AI strategy shaped by rigorous research and executed with real-world pragmatism, we’d be glad to partner with you.

‍

‍

The bottom line: AI itself is not a strategy – building the right system is. As you plan for 2026, focus on the human+AI operating model of your business. The companies that thrive will be those that marry the latest AI capabilities with process redesign, workforce enablement, and a clear vision for transformation. The technology is increasingly ready; the winners will be those who are ready to use it fully.

‍

Sources:

The state of AI in 2025: Agents, innovation, and transformation (McKinsey, 2025)

‍

Ground your workforce AI strategy in human experience (Forrester, 2025)

‍

‍State of AI: An Empirical 100 Trillion Token Study with OpenRouter

‍

10 AI Predictions For 2026: Top Experts Share New Trends (Forbes, 2025)

‍

Why 88% of Companies Use AI but Only 6% See Real Results: What McKinsey’s Research Really Tells Us | Best of Digital Transformation

‍