Ten industries, one card each: a direct answer, the Fit Matrix score row, a two-paragraph overview, a worked example you can copy today, two adjacent use cases, and the starter use case to run this quarter. Read in order or use the chip strip to jump.
Healthcare
AI solutions for healthcare
Healthcare uses AI for clinical decision support, medical imaging, and patient-facing assistants — with FDA-cleared models leading deployment. Adoption sits at 62% of healthcare organisations; agent-readiness lags at 18% in production.
Healthcare is mid-pack on AI maturity (62% adoption per Medha Cloud) and bottom-tier on agent-readiness (18% per Joget's Gartner/IDC analysis), because regulatory drag and patient-safety review extend every deployment cycle by months. The result: most production work is decision support, not decision automation.
ROI velocity is solid (5.8x in 14 months — McKinsey, applied across industries; healthcare hits that range when paired with workflow-anchored use cases). The fastest wins are documentation, summarisation, prior-authorisation triage, and operational analytics — not diagnostic AI, which is still gated by FDA clearance and clinical-trial evidence.
Worked example
- Role
- Clinical operations lead at a 200-bed hospital
- Task
- Summarise 30 patient progress notes from yesterday's rounds for clinician review
- Workflow
- Paste de-identified notes → prompt: 'Summarise each note as: patient ID, key finding, follow-up.' Use a long-context model (Claude 4.5 Sonnet or Gemini 2.5 Pro). Pipe the structured output into the team's huddle doc.
- Result
- 10 minutes for 30 notes versus ~90 minutes manual — a 9× speed-up without touching clinical judgement.
Adjacent use cases
- Medical imaging triage — FDA-cleared computer-vision models flag urgent findings for radiologist review.
- Prior-authorisation drafting — generate first-pass appeal letters from EHR notes, edited by a human before submission.
Start here this quarterClinical-note summarisation for clinician review.
Financial services
AI solutions for financial services and banking
Banking leads agent deployment at 47% in production (Joget/Gartner-IDC) — fraud detection, KYC automation, and personalised advisory drive ROI. Adoption sits at 79% across financial services, second only to technology.
Financial services is the most agent-ready non-tech vertical in 2026. Banking and insurance together hit 47% production agent deployment (Joget summarising Gartner/IDC) — well ahead of healthcare and government at 18%. The combination of well-structured transactional data, clear compliance gates, and direct revenue tie-in makes the business case unambiguous.
AI-driven hyper-personalisation drives up to 92% higher digital engagement and 10–25% revenue growth from tailored offers (TTMS, 2026). Fraud, KYC, claims triage, and AML monitoring are the four highest-ROI starter workloads — each pairs measurable risk-reduction with auditable model output.
Worked example
- Role
- Anti-fraud analyst at a mid-market commercial bank
- Task
- Triage 500 flagged transactions overnight before the analyst team arrives
- Workflow
- Stream flagged transactions into a structured prompt: 'For each row, classify risk (low/med/high), cite the top 2 features driving the score, suggest next action.' Route high-risk to human review queue; auto-clear low-risk per policy.
- Result
- Analysts open the day with a triaged queue instead of 500 raw rows — typical 4–6 hours of triage compresses to a 30-minute review pass.
Adjacent use cases
- KYC document review — pull entity, address, and beneficial-owner fields from passports, utility bills, and incorporation docs.
- Personalised advisory drafts — generate first-pass investment-summary notes for advisor review before client meetings.
Start here this quarterFraud + KYC agents in regulated workflows.
Manufacturing
AI solutions for manufacturing
Manufacturing uses AI for predictive maintenance, computer-vision quality control, and digital-twin simulation. Pair AI with IIoT sensor data and McKinsey's 5.8x ROI in 14 months benchmark consistently holds for production deployments.
Manufacturing's AI story is bimodal: tier-one OEMs are deep on predictive maintenance and computer-vision QA, while most mid-market plants are still on horizontal generative tools (writing SOPs, drafting RFQs). The bridge is workflow-specific — a single high-cost downtime line, instrumented with IIoT sensors, can deliver McKinsey's 5.8x ROI inside one fiscal year.
Agent-readiness is mid-pack: production environments demand deterministic behaviour and clear human-in-the-loop checkpoints. Expect supervised rather than autonomous deployments through 2026, with agentic patterns appearing first in maintenance scheduling and supplier-document triage.
Worked example
- Role
- Plant reliability engineer at a Tier-2 automotive supplier
- Task
- Predict bearing failures on the press line 48 hours before they shut production
- Workflow
- Stream vibration and temperature sensor data into a time-series anomaly model → call a generative model with the anomaly report and the last 30 days of maintenance log to draft the work order, including parts, technician skill level, and downtime window. Engineer approves and dispatches.
- Result
- Unplanned downtime drops 30–60% on instrumented lines (industry-typical range) — single-line ROI usually clears 5x inside year one.
Adjacent use cases
- Computer-vision quality — flag surface defects on a moving line at human-eye accuracy, route to QC for confirmation.
- Digital-twin simulation — pair generative AI with a process simulator to test changeover sequences without touching the floor.
Start here this quarterPredictive maintenance on the highest-cost downtime line.
Retail & e-commerce
AI solutions for retail and e-commerce
Retail uses AI for hyper-personalisation, inventory forecasting, and visual search — with up to 92% higher engagement from tailored product feeds (TTMS, 2026). Adoption sits at 53% and is the fastest-growing among non-tech verticals.
Retail is the highest ROI-velocity vertical outside tech and finance. The reason is structural: retail has the densest first-party behavioural data of any industry, and every personalisation lift maps directly to conversion. TTMS reports up to 92% higher digital engagement and 10–25% revenue growth from AI-tailored offers in 2026.
Agent-readiness is rising fast — chat-based shopping assistants and post-purchase support agents are moving from pilot to production at most mid-and-up retailers. The blocker is now content operations, not model capability: an agent is only as good as the product taxonomy behind it.
Worked example
- Role
- E-commerce merchandising manager at a $50M DTC apparel brand
- Task
- Generate 200 personalised product-recommendation feeds for VIP customers ahead of seasonal launch
- Workflow
- Pull last-12-month purchase history + browse logs per VIP → prompt a generative model with the segment summary, the new-season catalogue, and the brand voice guide: 'Draft a 3-product recommendation block per customer, with one-line copy each.' Merchandiser reviews and approves in batches.
- Result
- 200 personalised emails in 90 minutes versus a full-week merch sprint — pilot-typical conversion lift of 15–30% per personalised cohort.
Adjacent use cases
- Visual search — let shoppers upload an image, return the closest catalogue matches with confidence ranking.
- Inventory forecasting — combine point-of-sale, weather, and event data for next-week SKU-level demand.
Start here this quarterPersonalised product feeds tied to first-party purchase history.
Legal
AI solutions for legal
Legal uses AI for contract review, e-discovery, and research — gated by privilege concerns but accelerating fastest in mid-market firms. Mid-market is moving faster than Big Law because the partner-protection layer is thinner.
Legal's adoption curve in 2026 looks more like SaaS than law: mid-market firms lead deployment because their partner-protection ceremony is shorter than Big Law's, and their billable economics push faster cycle times. Contract review, discovery triage, and matter-research synthesis are the three highest-ROI workloads.
Agent-readiness is constrained by privilege and accuracy thresholds — a single hallucinated citation can kill a deployment. The pattern that works: long-context summarisation with explicit source-page citation, gated by human review for any output that leaves the firm.
Worked example
- Role
- Mid-market commercial litigation associate at a 60-attorney firm
- Task
- Redline a 40-page master services agreement against the firm's standard clause library before partner review
- Workflow
- Upload the MSA and the firm's standard clause library → prompt a long-context model: 'For each non-standard clause, flag the deviation, cite the firm-standard equivalent, and propose a redline that preserves the client's intent.' Associate reviews each flag, accepts/edits, then sends to partner.
- Result
- First-pass redline in 45 minutes instead of 4–6 hours — partner gets a cleaner draft, associate keeps the final-edit judgement.
Adjacent use cases
- Discovery triage — cluster and tag millions of documents by relevance and privilege before manual review.
- Matter research — generate a sourced memo on a narrow legal question with each citation linked to a specific page.
Start here this quarterContract redlining against firm-standard clause libraries.
Education
AI solutions for education
Education uses AI for tutoring, grading, and content generation. Adoption lags at 34% (Medha Cloud, 2026) due to procurement cycles, but pilot velocity is up sharply — and SMB-style tutoring tools are the gateway most institutions take.
Education is the largest under-served opportunity on this matrix. Adoption sits at 34% (Medha Cloud), well below tech, finance, and healthcare, but it is climbing fastest in K-12 and community colleges where procurement cycles have shortened.
ROI is real but slower than retail or finance because the value flows to time-saved-by-teacher and learning-lift-per-student rather than direct revenue. Tutoring assistants, grading copilots, and assignment-feedback generators are the three highest-yielding starter workloads.
Worked example
- Role
- High-school AP-Calculus teacher with 120 students across 4 sections
- Task
- Grade weekly problem sets and write personalised feedback for 120 submissions
- Workflow
- Scan submissions, OCR to text → prompt a model with the rubric and student work: 'Score by rubric, write one sentence of constructive feedback per problem, flag any student showing repeated conceptual errors.' Teacher reviews flagged students individually.
- Result
- Friday-night grading drops from 4 hours to 45 minutes; the flagged-students list becomes the Monday office-hours agenda.
Adjacent use cases
- Tutoring assistant — Socratic dialogue on a specific topic, paced to the student's prior responses.
- Curriculum generation — draft a 2-week unit plan from a standards document, edited by the instructor.
Start here this quarterTutoring + grading assist on standardised assignments.
Logistics & supply chain
AI solutions for logistics and supply chain
Logistics uses AI for route optimisation, demand forecasting, and exception management — with measurable on-time-delivery lift. The fastest-ROI workload is route optimisation tied to real-time traffic and weather.
Logistics' AI maturity is mid-pack and rising. The leading edge is exception-management agents — software that detects a shipment delay, drafts a customer notification, reroutes inventory from the next-best DC, and updates the ERP without human keystrokes.
Demand forecasting and route optimisation are the bread-and-butter wins. McKinsey's 5.8x in 14 months benchmark holds reliably when the AI is anchored to a single P&L-relevant metric (on-time-in-full, days-of-inventory, or cost-per-mile).
Worked example
- Role
- Transportation planner at a regional grocery distributor
- Task
- Re-plan tomorrow's 80-stop urban route after a major bridge closure
- Workflow
- Feed current route plan + closure data + driver-shift constraints into an optimisation model → call a generative model on the optimiser output: 'Summarise the change for each affected driver, draft an SMS, and list any customer ETAs that now slip more than 30 minutes.' Planner reviews and sends.
- Result
- Re-plan in 15 minutes versus 90 minutes manual; affected-customer notifications go out before drivers leave the yard.
Adjacent use cases
- Demand forecasting — combine POS, weather, and promotional data for SKU-level next-week orders.
- Carrier-document automation — extract BOL, POD, and invoice fields, reconcile against ERP records.
Start here this quarterRoute + demand forecasting on the highest-volume lane.
Professional services
AI solutions for professional services (consulting, accounting, marketing)
Professional services see the fastest ROI of any non-tech vertical. Research synthesis, drafting, and proposal generation are the highest-yield workloads — and the multiplier compounds because consultants bill against time saved.
Professional services is the dark horse of the matrix. Maturity is high (4/5) because individual consultants adopted generative tools faster than their firms could buy enterprise versions. ROI velocity tops the chart outside tech — every billable hour saved on research, drafting, or proposal work is a directly recapturable margin.
Agent-readiness is also strong (4/5) because the workflows are document-in, document-out, with clear human review gates. The pattern that works: every consultant gets multi-model access, every deliverable carries an AI-assist tag in the audit trail, and review checkpoints are explicit.
Worked example
- Role
- Senior consultant at a mid-tier strategy firm
- Task
- Build a 30-page client deck for tomorrow's steering committee from a week of interviews
- Workflow
- Upload interview transcripts + the prior steering deck → prompt a long-context model: 'Synthesise the interviews into 5 themes, propose a deck outline that maps each theme to a decision, and draft speaker notes.' Consultant edits, then asks an image model for chart prototypes.
- Result
- First draft in 90 minutes instead of a full-day deck sprint — partner review now happens the night before, not the morning of.
Adjacent use cases
- Proposal generation — pull case-study selections, draft scope, and price-point options from the firm's prior-engagement library.
- Research synthesis — turn 50 industry articles into a single sourced briefing per client industry per week.
Start here this quarterResearch + proposal drafting on every active opportunity.
Government
AI solutions for government and public sector
Government uses AI for citizen services, document processing, and benefits adjudication — but trails the private sector at 18% production agent deployment (Joget/Gartner-IDC) due to procurement and trust gates.
Government's AI maturity in 2026 is the lowest on this matrix. Procurement cycles, security clearance overhead, and public-trust requirements multiply every deployment timeline by 2–4× versus private sector. Agent-readiness is similarly low at 18% production deployment.
The use cases that do land are document-heavy and citizen-facing: benefit application triage, FOIA response drafting, permit-review acceleration, and tier-1 contact-centre assist. ROI is real but measured in citizen wait-time reduction rather than revenue.
Worked example
- Role
- Benefits-eligibility worker at a county human-services agency
- Task
- Triage a backlog of 400 SNAP applications for completeness before formal review
- Workflow
- Run each application through an OCR + extraction model → prompt a model with the policy checklist: 'For each application, flag missing documents, list which interview questions are still required, and rank by case-urgency factors.' Worker uses the triaged list to schedule outreach.
- Result
- Application turnaround drops from 21 days to 7 days for complete files; worker hours redirect from data entry to case-management calls.
Adjacent use cases
- FOIA response drafting — draft first-pass redactions on document requests for officer review.
- Citizen contact-centre assist — surface the right policy paragraph to the agent during the call instead of after.
Start here this quarterDocument-processing automation on the highest-volume backlog.
SMBs
AI solutions for small and mid-sized businesses (cross-industry)
SMBs benefit most from horizontal generative AI (chat, image, document, research) bundled under one subscription — the per-tool-stack math no longer pencils. One platform covers ~80% of cross-industry AI workflows without four to six vendors.
Most SMBs in 2026 are still paying separately for ChatGPT, Claude, an image generator, and a research tool. The math is straightforward: four $20-class subscriptions per seat clear $80/mo while one multi-model platform delivers the same coverage for $29.99/mo. The 5.8x McKinsey ROI shows up faster in SMBs because the implementation cost is effectively zero.
Agent-readiness for SMBs is mid-pack — the constraint isn't model capability, it's workflow design. The pattern that works: pick the single highest-volume document workflow (proposals, invoices, customer email), wrap it with a templated prompt and a human-review checkpoint, then expand from there.
Worked example
- Role
- Founder of a 12-person professional-services SMB
- Task
- Cut $360/mo of stacked AI subscriptions to one bill while keeping every model your team uses
- Workflow
- Inventory current seats (e.g., 4× ChatGPT Plus + 2× Claude Pro + 2× Midjourney + 1× research tool = $360/mo). Move every seat to one multi-model platform that includes the same labs plus image generation and research. Save the difference; standardise on shared workspaces.
- Result
- $360/mo collapses to roughly $120/mo on a multi-seat plan; team gets every model under one login.
Adjacent use cases
- Proposal-generation workflow — draft client proposals from a discovery-call transcript and your service-line catalogue.
- Operating-document drafting — turn meeting notes into SOPs, training docs, and onboarding checklists.
Start here this quarterConsolidate every individual ChatGPT/Claude/Midjourney seat into one platform.
The clearest SMB win is consolidating individual seats into one bill. Browse every ZeroTwo AI tool in one dashboard →