.png){kind=small}
Beyond the Pilot Trap: The AI Consultancy Guide For Enterprise AI Adoption in 2026
- Mark Evans MBA, CMgr FCMi
- 3 days ago
- 10 min read
Written by Mark Evans MBA
Executive summary
If you are still stuck in pilots in 2026, you are already behind.
The real advantage now sits in systems, governance, and context, not in tools.
AI agents will decide who is visible in your market. Being invisible to them is the same as not existing.
Boards need to treat context cost, ethics, and regulation as strategic issues, not technical noise.
For many UK leadership teams, the conversation about artificial intelligence has shifted from potential to imperative. Yet most organisations are still stuck in the early stages of adoption. Many are either experimenting or piloting. Very few have reached full, scaled deployment.
This article is written for UK boards, C suite and owners of firms who are already experimenting with AI and now need to move beyond pilots.
Heading into 2026, any business still caught in this extended experimentation phase should treat itself as officially behind. The window to establish structural competitive advantages is closing. Future success depends less on chasing the latest model and much more on adopting a systemic, enterprise-wide approach.
Pilots are not a strategy. A long list of experiments is often a sign of drift, not progress.
It is no longer enough to chase quick win pilots. When they are not anchored in a clear direction, they can do the organisation a disservice. The real question now is how to position the business to survive and thrive in markets that will be reshaped by autonomous agents, not whether to use AI at all.
I.The four mental shifts required for systemic adoption
Moving beyond experimentation requires a real shift in both strategy and culture. The transition from isolated experiments to enterprise deployment rests on four mental shifts, drawn from practical scaling work across the industry.
1. From tools to systems
The old technology playbook says: test a tool, prove the value, then scale it. AI does not move that way. Capabilities evolve in weeks rather than quarters.
Success now depends less on which vendor or model you choose and more on the systems you build around AI. That means the workflows, governance, integrations, security, and feedback loops that let AI become part of how work is actually done rather than a side experiment.
The strategic asset is not the model you choose. It is the operating system you build around it.
2. Thinking at a new velocity
The tools themselves do not stay still. New features appear on leading platforms at a pace that would have looked absurd even a few years ago. For some teams this feels like constant whiplash and creates a heavy operational burden.
Leaders cannot slow the pace of change, but they can change how the organisation responds. That means shorter decision cycles, clearer guardrails, and a willingness to treat AI as an ongoing capability, not a one off project.
3. Solutions from anywhere
AI is a cross cutting capability. A local improvement in one corner of the business can have value across the enterprise.
A marketing analyst who automates a reporting process may be building patterns and prompts that finance, operations, and customer service can reuse. Innovation can come from any employee, at any level, if there are channels to surface and scale it.
4. Compounding return on investment
AI impact should not be seen as isolated wins. Time savings, cost reductions, and new revenue streams are linked and compounding.
An automated workflow that saves two hours a week for one person may not look dramatic on its own. When the same approach is reused across functions, plugged into better data, and supported by stable governance, it becomes part of a compounding engine of benefit.
II. The practical path to scaling AI: a repeatable system
To move from pilots to real deployment, organisations need a repeatable, systematic process. A useful way to think about this is in four iterative phases.
Phase I: Setting the foundations (an ongoing process)
Foundations are not a one off checklist. They are a set of disciplines that run underneath everything else.
Leadership alignment Executives need to be involved early and be seen to use AI themselves. This is about visible leadership, not performative enthusiasm. At the same time, this must be a two way process. Leaders need to secure real buy in from employees and avoid overwhelming teams with constant tool changes and initiatives.
Governance that evolves Organisations that put robust, clear AI governance in place tend to move faster and with less internal friction. Governance is not a set of static rules. It needs to evolve as models, regulations, and use cases evolve.
A simple way for leadership teams to stay aligned is to be explicit about the primary intent of AI work. That can be framed around three anchors:
Productivity
Automation
Opportunity
If the intent is not clear, culture, communication, and investment all pull in different directions and paralysis follows.
Data improvement and context Data is still the loudest blocker to scaling AI. Typical problems include:
Fragmented data sources
Access friction and permissions
Heavy reliance on undocumented tribal knowledge
Leaders often underestimate the effort required to fix the data and the plumbing. Yet this is where a large share of value is won or lost.
Data work now ties directly into context engineering. That is the work of orchestrating memory, knowledge, and policy so that AI systems see the right information at the right time. New systems should be designed to be agent ready by default, not treated as stand alone silos.
Phase II: Creating AI fluency
Many organisations launch AI tools before building any meaningful level of skill. Adoption then stalls. Treating AI as a discipline that needs to be learned, practised, and rewarded changes that dynamic.
Building champions networks Champions networks are one of the most effective levers. The idea is simple. You identify and support people who are willing to explore AI in their own area, then help them translate that learning into the local context. Proficiency grows in the grain of the business, not in abstract training sessions alone.
Rewarding and enabling experimentation There needs to be a clear and formal way to share best practices, use cases, and strong prompts. Without this, every team reinvents the wheel.
There is also a basic contradiction. People are often too busy to learn the very thing that would give them back time. Leaders need to create official time away from day to day work for AI learning and experimentation. Organisations that support active AI communities tend to see much lower levels of shadow AI because staff feel they can explore openly.
If people only experiment with AI in secret, you do not have an innovation culture. You have a risk problem.
Phase III: Scoping and prioritisation
At this stage the goal is to build a clear system for capturing, evaluating, and prioritising opportunities.
Ideas should be collected through open channels, not just handed down from leadership or IT. Once collected, they can be assessed through a simple Value versus Effort matrix.
The high value, high effort quadrant usually contains the most important, enterprise changing use cases. These tend to be complex, cross functional and politically sensitive, but they are also where the long term advantage sits.
When prioritising, it is vital to design for reuse from the start. Reuse increases speed, reduces cost, and builds technical memory. You are not just solving one problem, you are building a pattern.
Phase IV: Building and scaling products
Here the objective is to turn good ideas into stable internal or external products. The guiding principle is iteration. AI systems can learn and improve. They do not need to be perfect on day one.
Structuring the right teams Systemic projects need cross functional teams. A healthy team mix will normally include:
Technical talent, such as engineers and AI specialists
Subject matter experts from the relevant business area
Data leads who understand where the information lives and how to clean it
An executive sponsor who can remove blockers
Unblocking the path The largest constraints on AI impact are often organisational. Access delays, unclear ownership, and slow approvals cause more damage than model limitations.
Governance frameworks can also become stale quite quickly. They should be reviewed and adjusted through this phase, not filed away once written.
III. The leader’s mandate: governance and value capture
Scaled AI adoption is now a leadership issue as much as a technical one. Governance and value capture need to sit at the core of the business, not at the edge.
The cost of context and governance
Context is expensive. Tokens, data retrieval, and infrastructure can account for a large share of AI operational costs within 18 months of scaling. Most boards are not yet used to thinking this way.
Context should be treated as a finite, budgeted resource. That means measuring and managing it, not hoping for the best. Simple board level metrics might include:
Token spend variance
Agent reset rate, how often the system loses the thread of a task or conversation
There is also a climate dimension. Tokens have a carbon shadow. Under the UK Sustainability Disclosure Standards, boards will need to consider the emissions profile of their AI workloads. Choosing a bloated architecture when leaner alternatives exist cuts against the principle of minimising emissions where it is technically feasible.
New architectures such as diffusion based language models and long context systems like Kimi K2 signal an important shift in the cost to performance curve. They point to a future where long context use is more affordable, but the underlying governance questions do not disappear.
Five strategic questions for the board
Autonomous AI agents, systems that can perceive, reason, and act on behalf of humans, are moving from theory into practice. If transaction costs fall towards zero, market structures will not look the same. Loyalty based on convenience will erode. If agents do not see you, you may as well not exist.
To shape their position before new structures harden, boards need to work through five strategic questions.
Disintermediation risk: are we visible where agents look?
If buyers use agents to source suppliers, can those agents find your business and read your structured data? If your information is not machine readable, you have already created a risk.
Agent strategy: walled garden or open agent?
Should you build your own proprietary agents, or ensure you are compatible with third party agents that customers already use? This is a decision about pricing power, not just technology. It will influence who controls the relationship with the end customer.
Ethical guardrails: what will we forbid?
Leaders need to decide now what behaviours are off limits for agents. Truthfulness, manipulation, data use, and consent all sit in this space. The simple test is whether you are comfortable with the means, not just the outcome.
Stakeholder impact: who benefits from efficiency gains?
When AI agents deliver efficiency, boards must decide how that value is shared. Employees, customers, suppliers, investors, and communities will all feel the impact. If all gains accrue to capital, social and political pressure will follow.
Regulatory preparedness: are we ready for transparency?
New rules on transparency, algorithmic audits, and liability are coming. AI agents may tacitly cooperate in ways that drive up prices or exclude competitors. That will attract regulatory scrutiny. Boards should be preparing for documentation, audit trails, and clear lines of responsibility now, not later.
If your AI strategy cannot be explained to a regulator, it probably cannot be defended to a customer.
Conclusion: leading in the agent era
AI agents mark a turning point. As with the internet or smartphones, there will be winners and losers. This time, autonomous decision makers are entering the economy. That forces a rethink of principles that have underpinned markets for decades.
The organisations that combine clear strategy with ethical foresight will be in the strongest position. Agentic commerce demands a rethink of how value is created, captured, and delivered. Companies that move early, with discipline, can redefine their sectors. Those that cling to old models or treat AI as a narrow, tactical tool risk being pushed to the edges as agents become the new gatekeepers.
Technology does not decide the outcome on its own. Values and choices do. The leaders most likely to succeed in an agent mediated economy will keep asking three simple questions:
Why are we using AI agents?
For whom are we creating value?
What values are we embedding in these systems?
The window to shape your position is still open. It will not stay open forever.
If you are a UK leadership team now stuck in pilots and want a structured outside view on foundations, governance and agent strategy, this is the work 360 Strategy does.
References
Acemoglu, D. (2024) The Simple Macroeconomics of AI. Cambridge, MA: MIT Shaping the Future of Work Initiative. Available at: https://economics.mit.edu/sites/default/files/2024-04/The%20Simple%20Macroeconomics%20of%20AI.pdf (Accessed: 1 December 2025).
Arya.ai (2024) State of Agentic AI Reliability 2024. Available at: https://www.arya.ai/research/state-of-agentic-ai-reliability-2024 (Accessed: 1st December 2025).
Calvano, E., Calzolari, G., Denicolò, V. and Pastorello, S. (2020) ‘Artificial Intelligence, Algorithmic Pricing, and Collusion’, American Economic Review, 110(10), pp. 3267–3297. Available at: https://www.aeaweb.org/articles?id=10.1257%2Faer.20190623 (Accessed: 1st December 2025).
Cambridge University Press (2025) Hunold, M. et al. ‘Algorithmic price recommendations and collusion: experimental evidence’, Journal of Industrial Organization Education (advance online). Available at: https://www.cambridge.org/core/services/aop-cambridge-core/content/view/CB651EEFF516B590F70D4A1447162FAF/S1386415725000092a.pdf/algorithmic-price-recommendations-and-collusion-experimental-evidence.pdf (Accessed: 1st December 2025).
Darden School of Business (2024) ‘Stakeholder Theory’, University of Virginia. Available at: https://www.darden.virginia.edu/stakeholder-theory (Accessed: 1st December 2025).
Department for Business, Energy & Industrial Strategy (2024) UK Sustainability Disclosure Standards: Implementation Guidance. Available at: https://www.gov.uk/government/publications/uk-sustainability-disclosure-standards (Accessed: 1st December 2025).
European Union (2024) Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (AI Act). Available at: https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng (Accessed: 1st December 2025).
Evans, M. (2025a) The Agent Readiness Playbook: Why Culture, Not Code, Defines Advantage [Strategic Article]. Glasgow: 360 Strategy.
Evans, M. (2025b) Agent Readiness, Part Two: Data and Technology That Actually Move [Strategic Article]. Glasgow: 360 Strategy.
Evans, M. (2025c) Beyond the Hype: The Hidden Costs of Context in AI Every Business Leader Must Know [Strategic Report]. Glasgow: 360 Strategy.
Evans, M. (2025d) Agent 2: Machine Speed Markets, AI Agents, Control of Demand, and the Fight to Stay Inside the Deal [Strategic Report]. Glasgow: 360 Strategy.
Inception Labs (2025) ‘Diffusion LLM // Finally it is working’. Available at: https://noailabs.medium.com/diffusion-llm-finally-it-is-working-4e19c0204f7c (Accessed: 1st December 2025).
Kant, I. (1785) Groundwork for the Metaphysic of Morals (Modern English translation). Available at: https://www.earlymoderntexts.com/assets/pdfs/kant1785.pdf (Accessed: 1st December 2025).
McKinsey & Company (2024a) The State of AI in Early 2024: Gen AI Adoption Sparks New Wave of Risk and Innovation. Available at: https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-in-2024 (Accessed: 1st December 2025).
McKinsey & Company (2025a) ‘The agentic commerce opportunity: How AI agents are ushering in a new era for consumers and merchants’. Available at: https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-agentic-commerce-opportunity-how-ai-agents-are-ushering-in-a-new-era-for-consumers-and-merchants (Accessed: 1st December 2025).
McKinsey & Company (2025b) The state of AI in 2025: Agents, innovation, and transformation [Global Survey]. November 2025.
Moonshot AI (2025) ‘Kimi K2 Technical Specifications’. Available at: https://kimi.moonshot.cn (Accessed: 1st December 2025).
OpenAI (2025) From Experiments to Deployments: A Practical Path to Scaling AIÂ [Guide]. Available at: https://cdn.openai.com/business-guides-and-resources/from-experiments-to-deployments_whitepaper_11-25.pdf (Accessed: 1st December 2025).
Shahidi, P., Rusak, G., Manning, B.S., Fradkin, A. and Horton, J.J. (2025) ‘The Coasean Singularity? Demand, Supply, and Market Design with AI Agents’, in The Economics of Transformative AI. Chicago, IL: University of Chicago Press (NBER chapter). Available at: https://www.nber.org/books-and-chapters/economics-transformative-ai/coasean-singularity-demand-supply-and-market-design-ai-agents (Accessed: 1st December 2025).
Turing Institute (2024) AI Infrastructure and Carbon Intensity: UK Enterprise Study. London: The Alan Turing Institute. Available at: https://www.turing.ac.uk/research/ai-infrastructure-carbon (Accessed: 1st December 2025).
UK Government (2023) AI regulation: a pro-innovation approach (White Paper). Available at: https://www.gov.uk/government/publications/ai-regulation-a-pro-innovation-approach/white-paper (Accessed: 1st December 2025).
Workplace Intelligence and Writer.com (2025) AI Adoption Survey 2025. Available at: https://www.workplaceintelligence.com/ai-adoption-survey-2025 (Accessed: 1st December 2025).