A Year That Changed Everything and Not Enough
The year that followed ChatGPT’s November 2022 launch has been extraordinary by the measure of AI capability development. GPT-4 in March 2023 raised the capability threshold substantially. Claude, Gemini, Llama, and a proliferating set of open and commercial models followed with capabilities that continue to expand. The commercial AI application ecosystem built on these foundation models has grown faster than any previous enterprise software category. The executive conversations about AI that were theoretical a year ago are now urgent.
By the measure of enterprise AI deployment delivering at-scale business outcomes, the year has been more modest. AI pilots have proliferated. Production deployments are growing. A small number of enterprises are generating meaningful, measurable business value from AI at scale. The majority of large enterprises have invested in AI exploration without yet reaching the production deployment scale that delivers the business outcomes the exploration was designed to enable.
The honest assessment is that most boards have a flawed picture of where their organisations actually are in this journey. The technology leaders who are briefing them tend to lead with the pilots that are progressing well and the capabilities that the AI tools have demonstrated, rather than with the organisational investments that determine whether the capability reaches production and the gaps that are preventing that transition.
The Genuine Progress: What Has Actually Changed
The genuine progress deserves to be named specifically. Productivity improvements for individual knowledge workers using AI assistance are real and measurable. Writing, research, code generation, summarisation, and analysis tasks are genuinely faster with AI assistance for most users, in ways that translate to measurable time savings. The enterprises that have deployed AI assistance tools broadly and provided training on effective use are seeing utilisation rates and productivity benefits that validate the investment.
A small number of use cases have reached production deployment at scale and are delivering measurable business outcomes. Customer support automation, where AI-assisted agents handle a significant proportion of customer interactions with high customer satisfaction scores, is the use case with the most consistent production evidence across industries. Code assistance for development teams, where AI-assisted code generation and review has measurably accelerated delivery velocity, is the second. Document processing and extraction, where AI systems handle high-volume document analysis tasks with accuracy approaching human performance, is the third.
These are real results. They are also narrower in scope and more limited in number than most board-level AI strategy presentations suggest.
The Persistent Gaps: What Has Not Changed
The foundational investments that determine whether enterprise AI delivers at scale have not followed the enthusiasm with the speed required. This is the part of the honest assessment that most technology leaders are not presenting to their boards with adequate clarity.
Data infrastructure remains the primary constraint. The AI use cases with the highest potential business value are those that require access to the organisation’s proprietary data: customer behaviour patterns, operational data, proprietary documents, historical performance. These use cases require data that is accessible, high quality, and governable. In most large enterprises, it is none of these things consistently. The data infrastructure investment that would unlock these use cases is understood to be necessary, is consistently approved in principle, and is consistently underfunded relative to the AI application investment it needs to precede.
Governance frameworks are incomplete at exactly the point where AI deployment is expanding fastest. The use cases that are in production are, for the most part, the lower-risk use cases where governance gaps have limited consequences. The use cases that are entering the pipeline include higher-risk applications where AI outputs affect consequential decisions: customer credit decisions, claims processing, personalised medical information, regulatory compliance assessments. Boards are approving these deployments without governance frameworks adequate to manage the liability exposure they create.
Skills development has not kept pace with deployment. AI tool adoption without skills development produces individual users who are using AI tools suboptimally and without the critical evaluation habits that prevent over-reliance on incorrect outputs. The structured AI skills development programmes that would systematically improve this across the enterprise have been deferred in favour of getting tools deployed.
What Boards Are Still Missing
The board conversation about AI that most large enterprises are having focuses on capability (what AI can do), investment (what we are spending), and aspiration (what we aim to achieve). The conversation that is not happening with adequate frequency focuses on readiness (are we actually prepared to deploy AI at the scale we are aspiring to), risk (what are the liability and operational exposures from current deployments), and honest progress (how far are we actually from the business outcomes the investment is designed to deliver).
Boards should be asking specific questions that most technology briefings do not answer. For each significant AI use case in development or production: who is accountable for the business outcome this AI is designed to improve, and what are they measuring? What governance framework governs the decisions this AI assists or automates, and who has approved it? What data infrastructure investment is required to take this from pilot to production, and is it funded?
The technology leader who provides crisp, honest answers to those questions is giving the board what it needs to govern AI investment responsibly. The one who provides capability demonstrations and aspiration narratives is not.
The Assessment That Sets Up the Next Year
The organisations that will look back on 2024 as a year in which enterprise AI delivered measurable business value at scale are the ones that enter it with a clear-eyed assessment of their current position. Not where they hope to be, but where they actually are: which AI capabilities are in production and delivering, which are in pilots that have not yet identified the path to production, and which are aspirational in the absence of the data and governance infrastructure they require.
That assessment, shared honestly with the board, is the foundation for the prioritised investment decisions that close the gaps. The alternative is another year of pilot proliferation and production constraints, with a widening gap between the organisations that have made the foundational investments and those that have not.
The ChatGPT moment was a year ago. The enterprise AI reckoning is now.