.png)
Lucent Search research | AI & Data Leadership Report 2025–2026
Scaling AI from pilot to production
What needs to be in place for a promising AI solution to become a dependable part of the business?
A successful pilot provides evidence that an approach may work. Putting it into operational use introduces further requirements: integration with existing systems, changes to workflows, clear ownership and the capacity to maintain performance.
Lucent’s research found that scaling AI solutions and integrating them into existing environments were widespread challenges among the leaders surveyed.
For executive teams, the findings raise questions about how projects are selected, what funding covers and who takes responsibility for delivery beyond the initial demonstration.
RESEARCH FINDINGS
What the research found
91% found scaling AI solutions challenging
37.6% of respondents rated scaling AI solutions as very challenging, while 53.8% found it moderately challenging.
A further 8.6% reported that scaling was not a challenge.
This finding describes respondents’ experience of difficulty. It does not mean that 91% of AI projects fail, or that the same proportion of organisations cannot move pilots into production.
89% found AI integration challenging
39.1% rated integration as very challenging, with a further 50% finding it moderately challenging.
Integration can involve connecting AI with existing systems, data and business workflows. The finding makes those connections a relevant consideration when assessing the resources and capability required for delivery.
79% found keeping up with technologies and tools challenging
39.8% rated keeping up with new technologies and tools as very challenging, while 38.7% found it moderately challenging.
For leadership teams, this raises a question about how to evaluate new developments while sustaining the systems and programmes already in use.
BUSINESS IMPLICATIONS
What this means for the business
Moving AI into operational use creates responsibilities that may extend beyond the team that developed it.
An operational function may need to change how work is allocated or decisions are made. Technology teams may need to support integration and reliability. Risk teams may need evidence that controls are appropriate. Finance will need to understand the ongoing cost and the value achieved.
These responsibilities affect the investment decision from the outset. A pilot budget may establish technical feasibility while leaving substantial work to be funded before the business can use the result.
The leadership question is whether that work has been anticipated, assigned and resourced. Without those commitments, a technically promising solution may reach the end of its pilot with no agreed route into everyday use.
ADOPTION
Rebecca’s perspective: establish the route into use early
I would want the leadership team to understand what happens if a pilot succeeds.
Who will use the result? What will change in their work? Which systems must it connect to? Who will maintain it, and what evidence will justify further investment?
The answers can develop as the organisation learns. Exploring them early helps expose dependencies that could otherwise remain hidden until significant time and money have been committed.
I would examine four areas.
The business process and its owner
Identify the process or decision the AI solution is intended to improve and the person accountable for that improvement.
The operational owner should help define the requirements, assess whether the solution is usable and establish how it will become part of the workflow.
This gives the development team a clearer understanding of the environment in which its work must perform.
Integration and production requirements
Establish which data, systems and technical capabilities are needed for operational use.
Depending on the application, this may include access controls, monitoring, reliability requirements, support arrangements and procedures for handling errors or interruptions.
These requirements should inform the delivery plan and investment case. They also help clarify which responsibilities sit with AI, data, engineering and technology operations.
Adoption and changes to work
Define what people will need to do differently and how the organisation will support that change.
Consider training, supervision, revised responsibilities and how users will raise concerns. Where AI informs a decision, establish who exercises judgement and who remains accountable.
Adoption needs an owner and sufficient operational capacity to make the change workable.
Funding and decisions about progression
Agree what evidence is required to continue, revise, pause or stop the initiative.
That evidence should consider business usefulness, operational readiness, risk and the cost of sustained use. A technically successful pilot may still need further development or may not justify expansion.
Clear decision points help leaders direct resources towards work with a credible route to value.
LEADERSHIP QUESTIONS
Questions for your leadership team
Use these questions when reviewing a pilot or a programme that is proving difficult to scale.
Who owns the business outcome?
Is the operational owner involved in defining and assessing the solution?
What must change for the solution to be used?
Have workflow, training and responsibility changes been identified?
Which dependencies remain unresolved?
Consider data access, system integration, engineering capacity and assurance.
Who will maintain performance after deployment?
Are support, monitoring and ongoing costs assigned?
What evidence will justify the next investment?
Are the criteria for progression clear to business and technical stakeholders?
Does the leadership team have the capability to deliver the transition?
Where is stronger experience in integration, operational delivery or change required?
HIRING IMPLICATIONS
What this means for leadership capability
The capability required to explore an AI opportunity may differ from the capability needed to embed it across an established organisation.
As work moves towards operational use, the leadership requirement may place greater weight on engineering delivery, integration, risk judgement, cross-functional coordination and changes to business processes.
Those capabilities may sit across several leaders. The appointment decision should reflect what is already present and where accountability or expertise is missing.
A stalled programme therefore deserves diagnosis before a new role is defined. The constraint may involve leadership capability, technical capacity, unresolved dependencies or insufficient commitment from the business functions expected to use the result.
Lucent’s Leadership Requirement Diagnostic examines these conditions alongside the outcomes the organisation needs. It helps define the experience, authority and remit required for the next phase of delivery.
EXPLORE RESEARCH
Explore related findings
AI ownership, accountability and executive sponsorship
How cross-functional commitments and decision rights support delivery.
AI investment and return on investment
Funding constraints and the evidence required to demonstrate business value.
The skills AI and data teams need to deliver
The technical, commercial and organisational capabilities required across the team.
Explore the full AI & Data Leadership Report →
ABOUT
About this research
This page draws on the organisational challenges section of Lucent Search’s AI & Data Leadership Survey Results 2025–2026, including questions about cross-departmental collaboration and organisational buy-in.
The findings reflect the participating leaders’ experiences at the time of the survey. They should be read alongside the respondent profile, question wording and methodology.
Research and interpretation by Rebecca Hastings, Founder of Lucent Search.
CONNECT
Reviewing the leadership needed for AI delivery?
If promising work is struggling to reach operational use, understanding the constraint will help clarify the leadership requirement.
Speak to Rebecca about what the business needs to achieve, where delivery is slowing and which capabilities are already in place.
Lucent Search · AI & Data Leadership Report 2025–2026