The future of AI will be decided by trust, not speed

Thomson Reuters CEO Steve Hasker argues the AI market is splitting between general-purpose tools and fiduciary-grade systems built for law, tax, and audit workflows. The piece emphasizes that trust, verifiability, accountability, and data ownership are the key differentiators for professional AI, rather than raw model speed. It also highlights the Trust in AI Alliance, involving Anthropic, AWS, Google Cloud, OpenAI, and Thomson Reuters, as an effort to establish shared principles for responsible AI.

Analysis

The market is underpricing the bifurcation between horizontal AI adoption and regulated workflow penetration. In fiduciary sectors, the binding constraint is not model quality but liability transfer: whoever owns the workflow, the audit trail, and the indemnifiable content layer captures the economic rent. That makes incumbents with embedded distribution and proprietary databases more valuable than frontier-model vendors whose products remain substitutable at the interface layer.

TRI is positioned to monetize a second-order shift: as generic copilots compress the cost of first-draft work, pricing power should migrate toward systems that reduce error rates and settlement risk, not just labor hours. The key implication is higher attachment rates in legal/tax/compliance bundles and lower churn, because the buyer is optimizing for defensibility and insurance value, not novelty. That can support multi-year ARR durability even if headline AI feature differentiation narrows.

The contrarian setup is that AI enthusiasm may be misread as a threat to incumbents when it is actually a selection mechanism. Open models likely commoditize low-stakes research, but they also raise the penalty for bad outputs, increasing willingness to pay for curated authority, provenance, and workflow integration. The main risk is that large legal-tech platforms or cloud vendors bundle similar functionality aggressively, forcing TRI to defend through content quality and trust rather than pure software breadth.

Catalyst-wise, the next 6-18 months matter more than the next quarter: expect buyers to pilot in narrow workflows first, then expand only after internal risk committees approve. Any high-profile hallucination-related litigation, regulatory scrutiny, or data-residency controversy would accelerate procurement of governed systems and strengthen the moat. Conversely, if frontier models materially reduce hallucination rates without compromising provenance, the premium on specialized workflow layers could compress.

Key Decisions for Investors

• Long TRI on a 6-12 month horizon; use any AI-related pullback to add, targeting rerating as investors price in higher retention and workflow monetization. Risk/reward favors the long side if trust-driven ARR expands faster than consensus.
• Pair trade: long TRI / short a basket of generic AI workflow beneficiaries with weak proprietary content or low switching costs (e.g., software names exposed to commoditized copilots). The thesis is that regulated verticals will take share from horizontal tool vendors over the next 12 months.
• Buy TRI call spreads 9-12 months out to express upside from AI monetization while capping premium spend; best if the market remains skeptical and management proves attach-rate expansion in professional workflows.
• Short-dated hedge: if TRI rallies sharply on AI narrative compression, fade with tight stops—near-term upside is likely to be multiple expansion, but the real fundamental inflection should be validated over several quarters.