From workflow automation to product reinvention, here’s where the real money shows up—and how to capture it.

Automate the Costly, Not the Flashy
The fastest path to ROI is unromantic: improve workflows that already have baselines. Claims triage, tier-1 support, invoice matching, collections, and QA for contact centers are full of repetitive steps with high labor minutes. Field studies show double-digit productivity gains when agents use copilots, with the largest improvements among new hires; build that into your 2026 plan with human-in-the-loop guardrails and weekly scorecards.

Reinvent Products with Agentic Features
In 2026, buyers expect software that can read screens, browse, run code, and complete tasks—“computer use” is shifting from demo to default. Package that capability inside existing SKUs (a research agent inside CRM; a reconciliation agent inside ERP), or sell new add-ons priced per seat or per action. Roadmaps from leading labs emphasize long-context reasoning and tool use, enabling agents that carry projects end-to-end rather than fire one-off responses. Design pricing to reflect work finished, not just drafts.

Monetize Proprietary Data—Safely
Your proprietary data remains the differentiator. Turn it into retrieval-augmented copilots for customers, premium analytics, alerts, or monitoring. Long context windows—now reaching a million tokens—reduce glue code and let users drop entire document rooms into a single run, unlocking “upload + ask” for regulated industries. Keep inference where the data lives (cloud for cloud data; plant or branch edge for local systems) and publish clear governance, access logs, and model/system cards.

Optimize Infrastructure Spend
Profit also comes from avoiding waste. Use elastic AI datacenters for bursty training, then shift steady inference to owned or colocated GPUs when utilization is predictable. IDC expects AI infrastructure spending to remain on a steep trajectory through the decade; treat compute like a portfolio—mix commitments with on-demand while watching unit economics. On-device frameworks and small multimodal models move everyday assistance off the cloud, cutting latency and cost for high-volume tasks.

Sell Outcomes, Not Tokens
The companies that bank the biggest gains don’t invoice for prompts or GPU-hours; they price business impact. Tie contracts to churn saved, revenue lifted, minutes cut, or cases resolved. Real-world telco programs show generative models predicting call intent, routing to the right agents, and trimming visit time—value business owners will pay for. Wrap it all with an “AI success kit” that includes playbooks, governance, and change-management.

What the 2026 Roadmap Looks Like
• Tool-Using Agents: Vendor SDKs are converging on multi-step agents that can browse, run Python, use files, and operate UIs. Build integration hooks now so your app becomes the canvas where agents work.
• Long Context as a Feature: Plan capabilities that assume corpus-scale inputs in a single run; reserve bespoke retrieval for highly dynamic or huge datasets.
• On-Device Contributions: Apple’s Foundation Models framework and Microsoft’s Phi-4 family accelerate lightweight, private inference for everyday UX.
• Budget Growth Continues—But Scrutiny Rises: Boards are asking for P&L impact, not prototypes.

Put It Together
Bundle the five plays: pick three measurable workflows; ship one agentic feature customers will pay for; turn your proprietary data into a copilot; right-size your compute mix; and invoice for outcomes. That flight plan turns pilots into P&L in a single quarter and scales across business units company-wide, fast.

Execution Guardrails
Avoid the trap McKinsey warns about—spending big with thin returns because the organization wasn’t rewired for AI. Stand up a product owner for each use case, connect AI to systems of record, and measure impact weekly. Treat prompt catalogs, safety reviews, and fine-tuning as operating practices, not experiments. If value stalls, pivot quickly to process steps with clearer baselines and shorter feedback loops.

Closing Thoughts
Profit in 2026 comes from being specific: one process, one product surface, one dataset at a time—then scaling what pays. Agents that act, context windows that fit the whole problem, and smart placement of compute make the economics work. The playbook is simple to say and hard to do: choose measurable workflows, monetize what only you know, and price outcomes over activity. Track, iterate, and compound small wins quarter after quarter, relentlessly.

References

• McKinsey — “The economic potential of generative AI” — https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier

• IDC — “Agentic AI to Dominate IT Budget Expansion Over Next Five Years” — https://my.idc.com/getdoc.jsp?containerId=prUS53765225

• Google DeepMind — “Gemini 2.5: Our newest Gemini model with thinking” — https://blog.google/technology/google-deepmind/gemini-model-thinking-updates-march-2025/

• Apple Newsroom — “Foundation Models framework unlocks new intelligent app experiences” — https://www.apple.com/newsroom/2025/09/apples-foundation-models-framework-unlocks-new-intelligent-app-experiences/

• Reuters — “Verizon uses GenAI to improve customer loyalty” — https://www.reuters.com/technology/artificial-intelligence/verizon-uses-genai-improve-customer-loyalty-2024-06-18/

Serge Boudreaux – AI Hardware Technologies
Montreal, Quebec

Peter Jonathan Wilcheck – Co-Editor
Miami, Florida