How AI Is Rewiring the Insights Industry: From Experimentation to Integration - Part 2

Part 1 of this two-part article, based on Altera Strategy’s ‘Future of Insights and AI Adoption 2026-29’ study, examined the scope and nature of Insights functions' and agencies' AI adoption and the benefits of this new technology.

Part 2 examines the barriers and challenges presented by AI and the planning required as the industry moves beyond experimentation to more integrated and transformative applications. This study is based on 43 interviews with corporate, agency and vendor leaders and analysis of 190 industry sources.

Challenges on the Path to Insights AI Adoption

A consistent and candid finding across corporate and agency interviews alike was that planning and management of Insights AI adoption remain nascent, and governance has not kept pace with usage. This is also true for organisations generally today and is also the case within many Insights functions and agencies. This is one of several challenges and barriers slowing AI adoption and the wider transition.

AI-Specific Risks

The report identifies four key risk categories that exist broadly, are relevant for Insights, and were reported by the study participants:

• Data breaches: uploading commercial or customer data to third-party tools

• Hallucinations and erroneous outputs: often difficult to detect

• Data homogenisation: AI-enabled analysis flattening data and losing nuance

• Weak usage: novice users accepting AI outputs without critical review

Data ‘Out of Control’

Participants also shared examples of unvetted, AI-generated data reaching and influencing business stakeholders without appropriate context or scrutiny. For some participants, this issue of data ‘taking on a life of its own’ had also proven a barrier to the adoption of knowledge management and their efforts to ‘democratise’ Insights data. (See also Knit's recent survey of US insights professionals, which found 92% had seen cases of AI-related risks involving non-researchers use of data).

Planning and Governance Maturity

Planning, governance and resourcing AI adoption is found to be a necessary priority for Insights functions and agencies, and should not be left to enterprise-wide initiatives or allowed to evolve organically. A corporate insights director described how ‘usage is overtaking governance’, with users adopting tools without oversight and controls arriving later. In response, a phased rollout of tools and capability development was seen as necessary.

Some ‘Enthusiasts’ whose enterprises are in the ‘Scaling’ stage of AI adoption are advancing with this work; but most are only beginning (see Part 1 for these segments). For these early adopters, the work of planning, preparation, skills and capability development is essential when moving beyond experimentation towards scaled and integrated AI use. One research vendor encouraged users to think of ‘prompting as the new briefing’ and to be realistic about what AI and technology can achieve, working with the limitations.

ROI Measurement Lacking

Despite near-universal usage, investment into training and systems, and strong anecdotal evidence of time saved, measurable return on Insights AI investment remains elusive for many of the agency and restech leaders interviewed. One corporate participant described AI as ‘the oxygen we breathe’ but admitted that ROI tracking within their own organisation remained informal.

Measuring the ROI of AI investments is a challenge not unique to Insights, especially at this stage of experimentation and scaling adoption. Several factors have contributed to this: challenges measuring time saved (or spent on checking) and total business impact; semi-structured trials with weak goals and KPIs; and usage running ahead of planning, governance and measurement.

Insights executives expected that their AI tracking metrics and rigour would evolve and improve heading into 2027. There was a desire to move beyond measuring ‘time saved’ to wider KPIs linked to productivity, quality and the revenue growth associated with AI-enabled solutions. Ultimately, a significant P&L impact is expected given the substantial investments often being made, and some initial payoff is required.

For their part, practitioners acknowledged that AI was saving time, but some were reluctant to give up this efficiency because it was providing more ‘time to think’, helping improve quality or simply ensuring workloads remained in balance despite growing demand and speed. They expected greater rigour around KPIs to be required and also saw value in more holistic KPIs than simply ‘time saved’.

Turbulent Corporate Environment

Several corporate leads expressed concerns about AI-driven cost cutting and being asked to ‘do more with less’. Recent research from Mercer shows almost 100% of corporate C-level executives expect AI-driven restructuring in the next two years and headcount reductions as a result. Some leaders posited that in this context, proactively leading on the Insights AI transformation to deliver efficiencies and make a case for reinvestment is required, rather than a reactive approach.

Legacy Challenges and No Established Adoption Playbook

The majority of those interviewed were navigating their adoption ‘solo’ and without a set plan. Some were managing their adoption proactively, seeking out training and internal or external resources. In a time of transition and without an established playbook, there was also a common demand for increased engagement from partners and vendors to support this education, planning and implementation. Several participants expressed frustration that this was less forthcoming than they had hoped.

Several agencies and corporate functions had engaged AI consultants, partnered with existing vendors or internal AI transformation teams; some with positive results. Though given the highly specific nature of Insights practices, processes and data, consultants and partners from within the industry were generally considered to be better positioned without the steep learning curve external consultants may require. 

Corporate leads also spoke to legacy internal systems and data issues, complicating AI adoption or simply the challenging scale of the task when moving from experimentation towards scaled adoption. In several cases, existing initiatives on this front are now overlaid with AI transformation efforts, adding further complexity while increasing urgency.

In most cases, there is also a volume of tacit knowledge that sits with experienced practitioners that requires capture and documentation before more integrated or agentic AI can be implemented. Preparation of legacy knowledge and data in the form of past studies, proposals or data sets is also required to prepare for AI use, connection and automation or agentic use and can extend across hundreds or even thousands of projects and substantial effort is required.

Impact on Insights Talent

Many participants expressed concern about the potential impact of AI on the career paths of those new to the industry and their skills development; several leaders expressed concern about skills development and the risk of accepting '100%' AI outputs.

Challenges exist in slower entry-level hiring as workflow efficiencies grow, and diminished opportunities to learn the fundamentals of research and higher-level critical and strategic thinking. Some saw a potential ‘talent crunch’ in 3-5 years if investment in new talent and skills waned. In contrast, others saw that the current wave of change will offer opportunities to those who were ‘tech-savvy’ and seeking to learn and innovate.

Competing AI Deployment Models

The study identified a debate around the role of AI and LLMs in the future toolkit:

• The Horizontal AI Model: Generic LLMs, with custom-built insights context, prevailing as the dominant workflow tool, orchestrating insights tools and data via MCP connections; or

• The Vertical AI Model: AI embedded in Insights-specific platforms and solutions with purpose-built knowledge, context, processes, guardrails and outputs (or some mix of the two models).

In answer to this question, it was notable that participants further along the adoption curve were more likely to be combining specialist Insights solutions and less likely to be relying heavily upon LLMs. The horizontal model requires custom adaptation to Insight, whereas the vertical model is advantaged in that it arrives packaged and fit for purpose from Insights vendors. As one vendor noted, "more and more we will see specialist AI tools for specific industries”.

Building the Insights AI Adoption Roadmap

Given the range of areas yet to be resolved, it’s little wonder that many practitioners feel challenged. In this context, a transformation plan and roadmap are required to progress through the Insights AI transition from experimentation to scaling AI in the function or agency and achieve a strategic integration of AI, avoiding a fragmented, delayed approach. This roadmap should interlock with the wider Insights strategic plan, forming a key pillar and driver of transformation.

Insights AI Transformation Plan

• This plan should set out the vision for the future insights function and role of AI around the dual roles of strategic advisory and data orchestration. Define how AI will integrate with and support the wider insights program and business, along with the main strategies and stages of implementation.

• The vision and goals for this plan should extend beyond delivering efficiencies to examine how the function evolves towards the dual roles and an emerging vision of always-on, predictive, democratised, AI-enabled Insights, connected and automated into a wider enterprise ecosystem.

• Business Case: The case can be made for how AI-driven efficiencies in programs and workflows can be reinvested to substantially grow the reach, impact and ROI of insights investment.

• Scope: The plan should extend to cover governance, talent, technology and data infrastructure, with consideration of the role of enterprise systems, agencies and vendors.

• Governance: As stewards and producers of data, an Insights-specific AI governance policy covering goals, uses, data, tools, processes and oversight is required.

• The plan sets out the goals and strategy; the roadmap defines the stages and steps required to execute the plan and the major stages and milestones required to achieve it over the short and medium term of 2026-2029.

• Staged Adoption and Transformation: From experiments to workflow implementation in key programs and preparation for future automated and agentic applications. Preparation is required at each stage, especially in larger organisations, with a strong foundation of technology and data infrastructure.

• Management, Talent and Communication: Defined goals, KPIs, resourced programs and the management mechanisms are required at each stage to build capability and advance transformation with supporting skills development and change management. A staged, achievable approach and investment are required across key solutions, tools and talent to progress successfully through the stages of adoption, linking investment with outcomes measured by KPIs and ROI. Initial expectations of dramatic short-term gains in productivity are now replaced by more realistic gains achieved over several years of transformation and capability development as the technology and solutions also develop.

• Evolution: The roadmap should plan for internal capability development and assume that LLMs and insights-specific solutions will evolve over the next two to three years.  Related to this is the expectation that the corporate AI and data ecosystem will also evolve rapidly.

• Moving to Best Practice:  A short-term ‘quick-win’ can be the adoption of a consistent approach to the use of LLMs and AI-enabled tools to ensure quality outcomes. This includes consistent data sources, processes, prompt and context engineering and libraries, guardrails, outputs, KPIs and oversight.

A Roadmap for AI and Growing Insights’ Role and Influence

A key conclusion of the study is that the industry is at an inflexion point of real opportunity and risk, affecting most insights functions, agencies and vendors. The industry’s ability to transform as it transitions to AI-enabled solutions and workflows will determine whether Insights grows in impact, ROI and influence, or devolves towards data provision outpaced by rapid change in the wider enterprise and adjacent functions such as marketing, CX and analytics.

A defined AI transformation plan and roadmap can help corporate Insights functions set their path forward; evolving legacy roles towards the dual roles of strategic advisor delivering deep human insights; and orchestrator of the Insights data helping power the future agentic, automated enterprise decision and data ecosystem.

Agencies can build their plan and roadmap around their unique sector and domain expertise, renewed legacy methods and new data sources and solutions, while developing solution integration and implementation capabilities, aligned to clients’ evolving needs and the new ecosystem.

Insights has the opportunity to secure a central role in this ecosystem, complementing traditional strengths in strategic advisory with a new and growing stream of measurable ROI driven by data connections and activation. As transformation settles and AI becomes ubiquitous through 2029, the combination of advantaged human insight with improved data connections and activation can set the scene for a renewed role, growing ROI and a new phase of growth.

‘The Future of Insights and the AI Transition 2026-29’  is an in-depth, independent study and knowledge base comprising 180p and 256 perspectives from 43 insights leader interviews and 190 industry sources. More information can be found at: https://www.alterastrategy.com/services