What is your AI rollout resolution for 2026?

My working hypothesis is that a large fraction of everyday knowledge work is already “LLM-like” in practice. Most tasks do not require explicit symbolic reasoning or grounded interaction with physical reality, but consist of ingesting language tokens (emails, documents, meetings) and emitting transformed language tokens (summaries, decisions, replies), guided by learned heuristics, priors, and social loss functions internalised through training and experience.

I observe that humans spend much of their professional time operating as probabilistic language processors running on organic hardware, optimising for coherence, plausibility, alignment with rules and norms, and minimisation of social or organisational loss, rather than formal logical correctness.

My New Year’s resolution is therefore to push more aggressively for an AI-first strategy: systematically identifying work that is reducible to language-in / language-out transformations, and delegating it to LLM- or domain-specific language-model-based systems (DSLMs). Not because humans are obsolete, but because this frees human cognition for the minority of tasks that genuinely require grounding, causality, judgment under uncertainty, or ethical responsibility. Because my 2025 ex-post realisation is that this is where the metaphor breaks down and human agency still matters.

I am seeing people slow the roll for AI in sales in 2026 until the better grasp best practices. To many people over indexed on AI for sales thinking it would increase sales by reducing headcount. That is simply not the case. People are now understanding that AI is still simply an enhancement of the process and can improve performance of the human, not yet replace the human in sales.

For smaller transactional sales AI can replace humans a bit more easily. Not for complex sales where more than 2 decision makers are involved.