Beyond the Hype: An Expert's Unfiltered View on AI's Real Impact on ICT Market Research
Introduction: A Sector That Researches Itself
There is a certain meta-quality to conducting market research in the Information and Communications Technology sector. The very tools transforming research methodology—artificial intelligence, natural language processing, cloud-based analytics platforms, and real-time data pipelines—are products of the industry being studied. ICT researchers occupy a unique position: they are simultaneously observers of and participants in the digital transformation they seek to understand.
The global ICT market is enormous and expanding rapidly. According to IDC, global ICT spending reached $4.5 trillion in 2022 and is forecast to grow at a CAGR of 5.1% through 2026, driven by cloud computing, cybersecurity, 5G infrastructure rollout, and enterprise AI adoption. Within this vast landscape, market researchers face a fundamental challenge: the sector moves faster than traditional research cycles can accommodate. A B2B software competitive analysis that takes six months to complete may be describing a fundamentally different market by the time it reaches decision-makers.
This opinion piece draws on observed best practices, published industry research, and the evolving methodological discourse within the ICT research community to offer a frank assessment of where AI is genuinely transforming research practice—and where the promises remain unfulfilled.
Where AI Is Actually Delivering Value in ICT Research
Let's begin with what is genuinely working. In three specific areas of ICT market research, AI-powered tools are delivering measurable improvements over traditional methods: secondary research synthesis, sentiment analysis at scale, and predictive demand modeling.
Secondary research in ICT has always been hampered by volume. The sector generates an extraordinary quantity of published material—analyst reports from Gartner, IDC, Forrester, and IHS Markit; technical white papers; patent filings; regulatory submissions to bodies like the FCC and BEREC; GitHub repositories; and developer community forums. No human research team can monitor all of this comprehensively. AI-powered synthesis tools, including purpose-built platforms like AlphaSense and Hebbia, are now capable of ingesting thousands of documents and surfacing thematically relevant passages with a degree of accuracy that would have been implausible five years ago.
Sentiment analysis has similarly matured. Earlier generation tools applied crude positive/negative/neutral classifications that were almost useless in B2B technology contexts, where sentiment is nuanced and domain-specific. Current large language model (LLM)-based approaches—particularly those fine-tuned on technology industry corpora—can distinguish between a CIO's cautious optimism about a vendor's roadmap and their genuine enthusiasm for a competing platform. This level of nuance is beginning to make AI-powered sentiment analysis a credible complement to (though not a replacement for) qualitative research.
The Genuine Opportunity: AI's greatest value in ICT research is not replacing human judgment but dramatically expanding the information surface that human researchers can engage with meaningfully. The researchers who will thrive are those who use AI to read more, not those who use it to think less.
The Uncomfortable Limitations Nobody Is Talking About
The research technology vendor community has done an impressive job of marketing AI capabilities while obscuring their limitations. As a research community, we owe it to ourselves and our clients to be more honest about where AI-powered research approaches are currently falling short.
The first major limitation is hallucination risk in quantitative claims. Large language models have a well-documented tendency to generate plausible-sounding but fabricated statistics. In an industry like ICT, where precise market size figures, growth rates, and competitive share data inform significant capital allocation decisions, this is not a minor inconvenience—it is a potentially catastrophic failure mode. Any research process that uses generative AI to produce market size estimates without rigorous human validation is taking an unacceptable risk.
The second limitation is recency blindness. Most commercially available LLMs have training data cutoffs that leave them months or years behind current market reality. In a sector where a hyperscaler's pricing change, a major acquisition (such as Broadcom's acquisition of VMware in 2023), or a regulatory ruling can reshape competitive dynamics within weeks, a research tool that cannot access current information is of limited strategic value.
Third, and perhaps most importantly, AI tools perform poorly on primary qualitative research. The nuanced, probing conversations that an experienced qualitative researcher conducts with a CIO about their vendor selection process, their unspoken fears about platform lock-in, or their organization's hidden decision-making dynamics cannot be replicated by any current AI system. ICT buying decisions—particularly at the enterprise level—are deeply political, emotionally complex, and relationship-dependent. Understanding them requires human-to-human engagement.
The Methodological Priorities That ICT Researchers Are Getting Wrong
Beyond the AI debate, there are structural methodological weaknesses in ICT market research that deserve attention. The most significant is over-reliance on self-reported technology adoption data. Survey respondents in enterprise ICT consistently overstate their adoption of advanced technologies. Studies consistently find that actual deployment rates for technologies like AI/ML, edge computing, and zero-trust security architectures trail self-reported adoption rates by 20–40 percentage points.
The solution is not to stop conducting surveys but to complement them with behavioral data wherever possible. Usage telemetry from software vendors (where consent is appropriately obtained), IT asset management data from firms like Flexera and Snow Software, and hyperscaler revenue disclosures all provide behavioral signals that calibrate self-reported survey data.
A second weakness is insufficient attention to the buying group. ICT purchase decisions, particularly for infrastructure, security, and enterprise software, typically involve 6–10 stakeholders with different priorities and information needs. Research programs that survey only CIOs or IT directors are capturing a systematically incomplete picture. The Chief Financial Officer's perspective on TCO, the CISO's risk calculus, and the end-user's workflow impact are all essential inputs to a complete understanding of technology adoption dynamics.
'We keep designing research around the vendor's org chart rather than the customer's decision-making reality. Until we fix that, our findings will continue to underperform as predictors of actual purchase behavior.'
What Excellent ICT Market Research Looks Like in 2024
The ICT research programs that are generating genuine strategic value today share several characteristics. They are continuous rather than episodic—built on always-on data streams supplemented by periodic deep-dive primary research, rather than annual report cycles. They are multi-stakeholder by design—engineering enough research touchpoints to understand the full buying group, not just the most senior technology executive. They are behaviorally grounded—anchored in what customers actually do with technology, not just what they say they intend to do.
Leading technology companies including Microsoft, Salesforce, and SAP have built substantial internal research functions that combine these elements. Microsoft's Customer and Partner Experience (CPE) research team, for example, integrates telemetry data from hundreds of millions of users with qualitative interview programs and large-scale quantitative tracking studies to produce insights that genuinely inform product strategy.
For independent research firms and in-house teams without equivalent scale, the priority should be deliberate methodological pluralism—the disciplined combination of quantitative and qualitative methods, primary and secondary sources, behavioral and attitudinal data. No single method is adequate to the complexity of the ICT market. The researchers who are adding the most value are those who resist the temptation of methodological monoculture, whether that monoculture is survey-based or AI-powered.
A Call to Methodological Maturity
The ICT market research community is at an inflection point. The tools available to researchers have never been more powerful. The stakes—informing trillion-dollar technology investment decisions—have never been higher. And the risk of methodological complacency, dressed up as innovation, has never been more real. The path forward requires intellectual honesty about both the genuine promise and the current limitations of new research technologies, combined with a recommitment to the rigorous, multi-method practices that have always distinguished excellent research from expensive noise.