Sectors Making Strides in AI: Who’s Rewriting the Playbook in 2025
- Lanre Adeoye
- Sep 22
- 5 min read
Updated: Nov 10
What leading companies actually do, how incumbents are adapting, and whether old workflows survive
By Lanre Adeoye
Some industries have moved past pilots. An “AI layer” is speeding up old workflows, but the real shift is where AI is changing how work happens. Across sectors, companies that redesign workflows (not just add tools) see outcomes, while broad deployment of autonomous agents is still early. See recent findings from McKinsey’s State of AI and BCG’s AI at Work 2025, which put agent integration into broader workflows at roughly 13% today (slideshow).
Finance: from rules and rows to context and challenge
Before: Siloed data, static rules, manual checks; underwriting tied to legacy scorecards and hand-keyed documents.
Now:
Quantexa turns siloed records into linked context through entity resolution and graph analytics so investigators see relationships rather than rows.
Ocrolus converts PDFs and statements into decision-ready data with human-in-the-loop validation, shifting teams from rekeying to auditing exceptions.
Zest AI brings explainable, fairness-aware underwriting with adverse-action logic and documentation, so policy becomes model plus governance rather than a black box.
How incumbents are adapting: Banks are pushing model risk and lineage into BAU, and ops teams are moving from typing to triage. McKinsey’s research notes that workflow redesign is the differentiator for ROI.
Prediction:
12 - 24 months, context-led triage becomes standard in fincrime at large banks; doc-to-data automates most intake.
3 - 5 years, rules persist for audit trails and edge cases, but the primary path is contextual investigation plus governed models. Old workflow survival: low to medium.
Healthcare: from clerical drag to ambient, structured care
Before: Clinicians typed during visits or dictated later; documentation, coding, and summaries were stitched across tools; outreach was scripted and manual.
Now:
Abridge captures the clinical conversation and produces billable, EHR-integrated notes for review and sign-off, with Epic workflow detail.
Ambience Healthcare turns one capture into multiple outputs across notes, summaries, coding and CDI, and is now embedded in Epic’s Toolbox.
Hippocratic AI runs protocol-bound care agents for post-discharge outreach, with a live UHS deployment and clear escalation to humans.
How incumbents are adapting: Health systems roll out ambient documentation service-line by service-line, with revenue-cycle and compliance teams adding review and downgrade rules.
Prediction:
12 to 24 months, “review and adjust” notes become common in high-throughput clinics.
3 to 5 years, manual note creation becomes an exception outside complex inpatient and specialty care.
Old workflow survival: low in outpatient, medium in complex inpatient.
Education and enterprise knowledge: from content churn to copilots
Before: Teachers built materials from scratch after hours; corporate knowledge sat in static intranets; feedback and tutoring were scarce.
Now:
MagicSchool acts as a copilot for lesson plans, rubrics, differentiation and comms, with reported reach across most U.S. districts and 160 countries.
Sana shows how AI-native knowledge + learning + agents are becoming enterprise infrastructure.
Khan Academy’s Khanmigo provides guided tutoring and teacher assistance with responsible-use guardrails.
How incumbents are adapting: School systems are formalizing disclosure and integrity policies. Enterprises are moving from search and assemble to ask and act assistants with stronger access controls. BCG finds only 13% see agents integrated into broader workflows, which explains the emphasis on governance and redesign.
Prediction:
12 - 24 months, teacher prep shifts from creation to curation, and enterprise portals become assistant front doors.
3 - 5 years, static intranets fade; teachers rarely start from a blank page. Old workflow survival: low in enterprise knowledge, medium in pedagogy with strict local rules.
Manufacturing and logistics: from fixed automation to adaptable autonomy
Before: Warehouse robots needed SKU-by-SKU training; WMS rules were static; inspections were periodic and manual.
Now:
Covariant deploys an RFM-1 robotics foundation model so robots can pick varied SKUs on day one and learn across the fleet.
GreyOrange orchestrates people, robots and inventory in real time with GreyMatter, replacing rule chasing with control-room optimization.
Gecko Robotics combines climbing robots and AI to build a first-order condition data layer for critical assets, shifting maintenance from periodic checks to model-driven planning.
How incumbents are adapting: Operators are learning failure modes and dataset drift. Industrial IT and ops teams are forming joint policy and tuning groups. Sector outlooks anticipate broad warehouse automation by 2027, with ranges by methodology (SupplyChainBrain; Interact Analysis).
Prediction:
12 - 24 months, real-time orchestration becomes default in modern fulfillment centers.
3 - 5 years, SKU-specific training remains for edge cases; preventive maintenance flips to model-driven at scale. Old workflow survival: low in automated sites, medium in legacy facilities.
Creative, media and retail: from linear pipelines to AI studios
Before: Video and content pipelines were long and linear; brand checks happened late; localization was slow.
Now:
Runway compresses storyboard to cut with Gen-4 generation and editing, improving consistency across scenes and characters as covered by The Verge.
Synthesia enables script-to-screen production with consent-based avatars and governance controls.
Typeface turns static brand decks into a dynamic brand hub, with agents that enforce voice, visuals and policy across regions and channels.
How incumbents are adapting: Creative teams are building evaluation sets for taste, tone and safety. Legal and brand governance shift earlier in the pipeline. Surveys show that over 60% of marketers already use generative AI in some capacity, with adoption deepening as governance matures (Salesforce State of Marketing).
Prediction:
12 - 24 months, rapid pre-vis and internal content become the default.
3 - 5 years, high-craft productions keep traditional pipelines, but most internal and much marketing content runs on AI studios with human direction. Old workflow survival: medium for premium craft, low for routine content.
Defense and public sector: from platform-centric to mission autonomy
Before: Operators controlled each platform separately; sensor fusion was slow; autonomy was narrow and brittle in GPS- or comms-denied settings.
Now:
Anduril uses Lattice to let a single operator compose and supervise multi-domain autonomous missions.
Shield AI advances Hivemind autonomy for single and teamed aircraft in denied environments.
Helsing demonstrates AI-piloted flight trials with partners, while building software-defined mission capabilities and swarm concepts.
How incumbents are adapting: Programs are adding autonomy hand-off rules and assurance cases. Procurement is adding software and data readiness criteria. Operators are training for mission composition and supervision rather than stick-and-rudder tasks.
Prediction:
12 - 24 months, mission-level autonomy expands in pilots and limited deployments.
3 - 5 years, wider fielding tracks with assurance and procurement cycles, but the operating concept moves toward composition and supervision. Old workflow survival: medium for redundancy and contested settings.
What to watch across sectors
Old workflows do not vanish overnight. They will persist for edge cases, contingencies, and audits while the new path becomes the norm.
Redesign the work, not just the tools. The fastest gains come from re-platforming core loops such as investigation, documentation, orchestration, production, and command. See McKinsey’s State of AI hub.
Governance is the unlock for scale. Only about 13% of workers report agents integrated into everyday workflows. Access and training are the main choke points. See BCG’s AI at Work 2025.
Sources (short list)
About the Author
Lanre Adeoye is a talent and business operations leader with experience at the intersection of people, technology, and strategy. An MBA graduate of London Business School, she has helped startups and multinationals scale across regions through innovative approaches to recruitment, organizational design, and workforce transformation. Her work now explores how AI and emerging technologies are reshaping work, leadership, and venture growth across industries.
Say hello on LinkedIn or at lanre.a@workarena.co
Comments