Reinventors
Salesforce
Agentforce passed 1 billion dollars in run rate while AI handling of half its support interactions let Benioff cut support headcount from about 9,000 to 5,000.
What 120 companies disclose about AI. And what they leave out.
I scored 120 software, healthcare and life sciences companies on the six scored questions in the AI Value Gap self-assessment, using only what they publish: annual reports, earnings calls, press releases and product pages. Most of them say what AI is earning. Very few say what it saved, what happened to the capacity it freed up, or how they decide what to fund. Scored 6 to 23 August 2026. The full table and the data are below.
// The finding
Companies publish the revenue side of AI and keep the operating side to themselves.
114 of 120
say something about new revenue from AI. 92 describe AI revenue that is live or scaling.
37 of 120
give any evidence on whether AI savings met what they expected.
75 of 120
say nothing either way about what happened to the capacity AI freed up.
116 of 120
say nothing about how they evaluate and fund new AI revenue bets.
Across all 720 answers, 244 could not be scored either way because the company publishes nothing on the point. That is 34% of the evidence, and it is not spread evenly. It sits almost entirely on the cost and funding side.
The same six questions are used in the AI Value Gap self-assessment. An earlier self-reported survey on these questions was closed with 14 responses, too few to publish. None of its answers is used here, and nothing here was entered into it.
// Where disclosure stops
Each bar is one question across all 120 companies.
Evidenced: a source supports the answerAbsent: the company covers the topic but reports nothing on this pointUndisclosed: nothing either way
// The quadrants
Where a company lands depends on how you read its silence.
Each company is placed in one of the four AI Value Gap quadrants. 108 companies have enough evidence to place. The other 12 do not, and are left out of these counts.
Count silence as zero, and the largest group is Watching (56 of 108). Score only what is disclosed, and the largest group becomes Fragile Growth (58 of 108): AI revenue on the record, nothing to show on the operating side. The first count is the headline. The second sits beside it and is never quoted on its own.
Efficiency ↑New revenue →
19The Efficiency Trap
8Reinventors
56Watching
25Fragile Growth
108 companies
Efficiency ↑New revenue →
7The Efficiency Trap
28Reinventors
15Watching
58Fragile Growth
108 companies
// In practice
Five companies, in one line each.
Two that disclose both sides, which is rare, one that discloses only the revenue side, and two where little is live yet. Each line is drawn from the company's own filings and announcements.
Reinventors
ServiceNow
Runs about 450,000 of its own AI agents for a stated 100 million dollars of 2025 savings while Now Assist reached about 600 million dollars in AI contract value.
Fragile Growth
NICE
Grew AI ARR 66 percent to 328 million dollars with AI in every new seven-figure CXone deal, but its own internal AI use is a single unquantified line in the 20-F.
Watching
Mettler-Toledo
Runs a live AI knowledge base that lifts service engineers' first-fix rates, but sells no AI product and quantifies no saving.
Watching
Bio-Rad Laboratories
Frames AI as a 2026 feature for future platforms and credits its margin work to supply-chain execution, so nothing AI is yet live on revenue or its own operations.
// All 120 companies
The full table.
Efficiency and revenue are each scored out of 100 on the self-assessment's scales, with silence counted as zero. Where the evidence-only quadrant differs, it is shown underneath. A score reflects what the company had published by the date it was scored, not how well it uses AI. If you think a score is wrong, say so through the contact page, with the source.
Showing all 120 companies
| Sector | Quadrant | Confidence | |||
|---|---|---|---|---|---|
| 10x Genomics TXG10x Genomics treats AI as a demand driver for its sequencing and spatial platforms and discloses nothing about deploying AI in its own operations, so it fails the coverage gate. | Life science toolsUS | Not enough evidence | 0% | 17% | Low |
| Adobe ADBEAI-first ARR crossed 500 million dollars and tripled year over year in Q2 FY2026, yet Adobe discloses almost nothing about using AI to improve its own efficiency. | SoftwareUS | Fragile Growth | 0% | 58% | High |
| Agilent Technologies AReports hard operational AI wins, with 80 percent of engineering drawings AI-generated and custom GC design cycle times cut 75 percent, but far less AI-attributable new revenue, landing it in the Efficiency Trap. | Life science toolsUS | The Efficiency TrapEvidence only: Reinventors | 75% | 33% | Med |
| Alcidion Group ALCShips registered AI medical-device capabilities inside a platform sold under record FY26 contracts, but discloses no AI role in its own operations and no AI revenue split. | Healthcare ITAU | WatchingEvidence only: Fragile Growth | 0% | 33% | Med |
| AppLovin APPAppLovin's AXON AI product revenue grew about 700% and its e-commerce arm hit a $1bn run rate, on a workforce so lean it earns roughly $17m of EBITDA per employee, though the efficiency is only lightly disclosed. | Ad techUS | Fragile Growth | 17% | 50% | Med |
| Asana ASANExited fiscal 2026 with over 6 million dollars of AI Studio recurring revenue on consumption pricing and eight customers above 100k, but at under one percent of revenue and with no disclosed internal AI efficiency the efficiency axis stays empty. | B2B SaaSUS | WatchingEvidence only: Fragile Growth | 8% | 42% | Med |
| Atlassian TEAMCut about 1,600 jobs to self-fund AI and runs it across sales, HR, finance and legal, while monetising Rovo through credit-based collection upgrades, landing as an efficiency-led story with revenue not yet as fully evidenced. | B2B SaaSAU/US | The Efficiency TrapEvidence only: Reinventors | 50% | 42% | High |
| Autodesk ADSKCut roughly 16% of its workforce across two rounds while its CEO wrote that the changes are not an effort to replace people with AI, has one named internal AI deployment dating to a 2024 CIO interview, and lists no AI feature on its published Flex rate sheet. | Design softwareUS | WatchingEvidence only: Fragile Growth | 17% | 33% | Med |
| Avantor AVTRDiscloses its full Revival transformation programme across go-to-market, manufacturing and e-commerce with no AI in it, and reports no AI revenue. | Life science toolsUS | Watching | 0% | 0% | Low |
| Bio-Rad Laboratories BIOFrames AI as a 2026 feature for future platforms and credits its margin work to supply-chain execution, so nothing AI is yet live on revenue or its own operations. | Life science toolsUS | Watching | 0% | 8% | Low |
| Bio-Techne TECHSays it is deploying AI across the organisation and designing AI-optimised proteins on 50 years of data, but quantifies nothing and none of it maps to a disclosed revenue line. | Life science toolsUS | Watching | 17% | 25% | Low |
| Braze BRZEReports 5.7m dollars of AI Decisioning Studio revenue and hundreds of customers on its new AI agents, while gross margin fell partly because it added headcount to deliver them and it reports no internal AI use at all. | Marketing SaaSUS | WatchingEvidence only: Fragile Growth | 0% | 33% | High |
| Bruker BRKRTouts AI-driven semiconductor metrology and SciY lab-digitisation revenue growing over 20 percent, but discloses nothing about using AI for its own efficiency, so coverage is low. | Life science toolsUS | Not enough evidence | 0% | 33% | Low |
| Butterfly Network BFLYGrew software to 35 percent of 2025 revenue behind its new Compass AI platform and Embedded partnerships, but discloses nothing on using AI to run itself. | Medical imagingUS | Fragile Growth | 8% | 50% | Med |
| Bytes Technology BYITThe only company in the cohort to publish a measured internal AI gain, a productivity increase worth 26 full-time roles, with agents live across sales, licensing and managed services, yet it discloses no AI revenue line at all. | IT resellerUK | The Efficiency TrapEvidence only: Reinventors | 67% | 42% | Med |
| Canva privateRuns a firm-wide AI Discovery Week with over 90 percent of staff using AI weekly and a 30,000-hour agentic saving target, and gates premium AI features behind paid tiers, but discloses no realised savings and no AI revenue funding model. | Design SaaSAU | WatchingEvidence only: Reinventors | 42% | 42% | Med |
| Celonis privateSells operational-efficiency AI to others through its AgentC suite but discloses nothing about running AI in its own operations, so it fails the coverage gate. | Process miningDE | Not enough evidence | 0% | 33% | Low |
| Certara CERTReports agentic AI cutting some scientific-service delivery steps by up to 80 percent, but its AI-attributable software revenue is real yet only growing modestly, landing it in the Efficiency Trap. | Biosimulation softwareUS | The Efficiency TrapEvidence only: Reinventors | 50% | 33% | Med |
| Charles River Laboratories CRLSells the AI-powered Logica discovery platform under a partnering model and joined Lilly's TuneLab, but treats AI mainly as a future demand tailwind and its own AI use is one digital-pathology workflow. | CROUS | WatchingEvidence only: Fragile Growth | 25% | 42% | Med |
| Cloudflare NETCloudflare cut about 1,100 jobs, a fifth of its staff, after moving its own operations to AI, and names AI demand as a driver of raised 2026 revenue guidance. | Internet infrastructureUS | Reinventors | 83% | 50% | High |
| Computacenter CCCIs helping build the world's largest AI cluster on a record 7.1 billion pound backlog, yet its own AI use fills one page of the annual report and the results statement describes its 46 million pound systems programme without using the word AI. | IT servicesUK | Fragile Growth | 17% | 50% | High |
| Confluent CFLT (delisted)Delisted after IBM completed its acquisition in March 2026, leaving a final annual report that mentions AI twenty-one times and internal AI use not once, and a pricing page that carries no AI meter at all. | Data streamingUS | WatchingEvidence only: Fragile Growth | 0% | 33% | Med |
| Craneware CRWLaunched the Microsoft co-developed Trisus Assist AI product in March 2025 but discloses nothing about using AI in its own operations, so coverage is low. | Healthcare ITUK | Not enough evidence | 0% | 25% | Low |
| CrowdStrike CRWDCut about 5 percent of staff citing AI as a force multiplier that flattens the hiring curve while its Charlotte AI SOC agent grew about 85 percent quarter on quarter, the clearest both-axes case in the batch. | Cyber securityUS | Reinventors | 50% | 50% | High |
| Danaher DHRRuns AI through its Danaher Business System, ships AI diagnostics such as the Aperio AI store and named a Chief AI Officer in October 2025, but quantifies no AI-attributable savings or revenue. | Life science toolsUS | WatchingEvidence only: Fragile Growth | 25% | 33% | Med |
| Darktrace privateAn AI-native security vendor scaling its ActiveAI platform across cloud, OT and identity with nearly 10,000 customers, but since going private under Thoma Bravo it discloses nothing about AI in its own operations. | Cyber AIUK | Not enough evidence | 0% | 50% | Med |
| Dassault Systèmes DSYSells live generative-AI Virtual Companions and is rebuilding its pricing around outcome and consumption-based AI monetisation, but attributes none of its 30 percent operating margin to AI. | Engineering and life sci softwareFR | WatchingEvidence only: Fragile Growth | 0% | 42% | Med |
| Databricks privateIts AI products run at roughly a 1.7 billion dollar annual rate, about a quarter of revenue, while internal AI shows up only as engineering tooling and no efficiency savings are quantified. | Data and AI platformUS | Fragile Growth | 17% | 50% | Med |
| Datadog DDOGSells AI on a published credit meter, has opened AI model training as a new market and reports over 750 AI customers, yet mentions internal AI use in no earnings call or filing and sells a product to measure AI impact without publishing its own numbers. | ObservabilityUS | Fragile Growth | 17% | 50% | Med |
| Definitive Healthcare DHCut data integration time about 25 percent through automation in 2025 and plans to add GenAI to its Vue platform next quarter, but nothing AI-driven is yet live on revenue. | Healthcare dataUS | Watching | 17% | 8% | Med |
| Dexcom DXCMLaunched the first generative-AI platform in glucose biosensing to power its over-the-counter Stelo sensor into prediabetes, but discloses almost nothing on internal AI efficiency. | Connected medical devicesUS | WatchingEvidence only: Fragile Growth | 8% | 42% | Med |
| Docusign DOCUReports that 75% of new code shipped is AI-assisted and internal agents are reclaiming 300 to 600 hours a week, but credits its margin beat to four drivers none of which is AI and breaks out no AI-specific revenue behind the AI-native IAM label. | B2B SaaSUS | The Efficiency TrapEvidence only: Reinventors | 67% | 42% | Med |
| Doximity DOCSSells a paid clinical AI suite to 140 health systems and has opened AI search ad deals with top-20 pharma, and uses AI in its own engineering, but says AI revenue will be minimal this year and discloses no AI-attributable savings. | Digital healthUS | WatchingEvidence only: Fragile Growth | 17% | 42% | Med |
| Duolingo DUOLDuolingo went AI-first internally and cut its AI video-call cost from about $0.30 to under $0.01 per call, while its Duolingo Max AI tier reached roughly 1.1m subscribers. | EdtechUS | The Efficiency TrapEvidence only: Reinventors | 67% | 42% | Med |
| Dynatrace DTThe CFO names AI as a reason to moderate hiring in support, G&A and sales, and the company has committed 915m dollars to buy Arize, but its own documentation states the agentic AI layer is not licensed and its rate card carries no AI line at all. | ObservabilityUS/AT | WatchingEvidence only: Fragile Growth | 42% | 33% | Med |
| Elastic ESTCCut approximately 7% of its workforce sixteen days after filing a 10-K that reported no restructuring and no internal AI disclosure, citing working in an age of AI automation, while over 600 large customers now use its AI capabilities and no AI revenue figure has ever been published. | Search and observabilityNL/US | The Efficiency TrapEvidence only: Reinventors | 58% | 42% | High |
| EMIS Group privateIts EMIS Scribe ambient AI is in live GP pilots and on the NHS AVT registry, but as a private Optum subsidiary it discloses nothing about revenue or its own operations. | Healthcare ITUK | Not enough evidence | 0% | 25% | Low |
| Endava DAVAThe first Reinventor whose reinvention is shrinking it, with AI-driven work tripling to 15% of revenue while the CEO admits clients are harvesting more of the AI productivity gain than Endava would prefer. | Technology servicesUK/RO | Reinventors | 58% | 50% | High |
| EPAM Systems EPAMAI-native revenue passed $160 million a quarter, 11% of the total, while the same shift pushed clients away from task-based services and forced a guidance cut; no internal AI saving is measured and headcount keeps rising. | Technology servicesUS | Fragile Growth | 42% | 50% | High |
| Evolent Health EVHAutomates a third of previously manual prior authorisations at scale and calls Q2 a tipping point in its AI journey, but reports no AI-attributable new revenue. | Value-based careUS | The Efficiency Trap | 75% | 0% | Med |
| Exact Sciences EXASRuns a $150M productivity plan with no AI attached to it, while machine learning underpins its launching Cancerguard multi-cancer test, so its only AI-linked value is a new but early revenue product. | DiagnosticsUS | WatchingEvidence only: Fragile Growth | 0% | 33% | Low |
| Experian EXPNMore than 17,000 Experian staff use AI tools that cut model approval times up to 70%, alongside AI-native products such as Patient Access Curator, though its new-revenue funding approach is undisclosed. | Credit and data servicesUK/IE | The Efficiency TrapEvidence only: Reinventors | 67% | 42% | Med |
| Figma FIGJust opened AI credit monetisation on Figma Make with over 80 percent of larger customers consuming AI credits weekly, while its own AI use is an anecdotal engineering productivity claim confined to coding. | Design SaaSUS | WatchingEvidence only: Fragile Growth | 17% | 42% | Med |
| Freshworks FRSHSays AI is suffused across every part of the business, cut 11% of staff in an AI-led restructuring, and has over 7,000 customers paying for an AI SKU, but discloses no AI revenue figure and no funding model. | B2B SaaSUS/IN | Reinventors | 83% | 50% | High |
| GE HealthCare GEHCHolds 115 AI-enabled FDA authorizations and is scaling AI software and SaaS revenue toward roughly 1.8bn dollars by 2028, and uses generative AI in field service, but quantifies no AI-attributable internal savings. | Medical imagingUS | Fragile Growth | 25% | 50% | High |
| GitLab GTLBMandates daily AI use company-wide and publishes its own internal adoption shortfall against a 90% target, cut 14% of staff while stating in writing that this is not an AI cost-cutting exercise, and is the only company scored so far to disclose the bar new product investment must clear. | DevOpsUS | Reinventors | 50% | 58% | High |
| Globant GLOBCut headcount about 10% while its subscription AI business grew ARR 60% in a quarter to $52.8 million at margins ten points above traditional delivery, priced on output rather than hours. | Technology servicesLU/AR | Fragile Growth | 42% | 50% | Med |
| Grammarly privateSells live generative AI in paid tiers on around 700 million dollars of ARR and has rebuilt itself into a multi-product AI platform, but discloses nothing about using AI in its own operations, so coverage is low. | Writing AIUS | Not enough evidence | 0% | 33% | Low |
| Guardant Health GHUses its InfinityAI platform to speed regulatory submissions and sells AI-supported biopharma data at a record $61M, but attributes its Shield cost-per-test cuts to lab automation and sequencing, not AI. | DiagnosticsUS | WatchingEvidence only: Fragile Growth | 17% | 33% | Med |
| Halma HLMASells AI-based clinical decision support through PeriGen and rides AI data-centre demand in photonics, but discloses nothing about AI use in its own operations. | Safety and health technologyUK | Not enough evidence | 0% | 33% | Med |
| Health Catalyst HCATSells a live AI intelligence layer (Ignite Intelligence, Healthcare.AI) as its main growth driver, but its own efficiency programme is a restructuring and 9% headcount cut with no AI attached. | Healthcare dataUS | WatchingEvidence only: Fragile Growth | 0% | 33% | Med |
| Hims and Hers Health HIMSRuns the MedMatch AI platform to power personalised care and provider decisions, its core growing revenue engine, but discloses no discrete AI revenue line and no AI-attributable internal savings. | Digital healthUS | WatchingEvidence only: Fragile Growth | 17% | 33% | Med |
| HubSpot HUBSMonetises Breeze agents on outcome-based pricing at 50 cents per resolved conversation with agent adoption up 80 percent, and guides two points of AI-driven margin expansion, but names no specific internal deployment areas. | B2B SaaSUS | WatchingEvidence only: Fragile Growth | 50% | 42% | High |
| ICON plc ICLRCuts 30 percent from contract-negotiation time with AI and is rolling Claude and Copilot across delivery, but its AI revenue is still an intent to monetise data assets rather than a live line. | CROIE | The Efficiency Trap | 67% | 17% | Med |
| Illumina ILMNSells AI-embedded informatics such as DRAGEN and Connected Multiomics, but tells its operational-efficiency story, including a $100M cost programme, with no AI in it at all. | Life science toolsUS | WatchingEvidence only: Fragile Growth | 0% | 33% | Med |
| Intuit INTURuns done-for-you agents across products touching over 50 million transactions a week, but reports no quantified internal AI savings and no discrete AI revenue line. | SoftwareUS | WatchingEvidence only: Fragile Growth | 17% | 42% | Med |
| IQVIA IQVRuns 294 AI agents across 90 use cases and names AI in a better-than-expected margin gain, while its AI revenue is described as winning work now and contributing more directly later. | Health data and CROUS | The Efficiency TrapEvidence only: Reinventors | 50% | 33% | Med |
| iRhythm Technologies IRTCIts entire Zio revenue rests on a deep-learning ECG read that runs across its clinical operation, but it attributes no specific savings, freed capacity, or headcount effect to AI. | Cardiac diagnosticsUS | WatchingEvidence only: Reinventors | 25% | 42% | Med |
| Judges Scientific JDGA full FY2025 annual report contains a single AI mention, and it is about customers' machine learning applications, not anything Judges deploys itself. | Scientific instrumentsUK | Watching | 0% | 8% | Med |
| Kainos KNOSSells AI delivery to government and enterprise with a 45.8 million pound AI revenue line and gives every employee a Copilot licence, yet grew its workforce 21% and reports no internal saving. | IT services and softwareUK | Fragile Growth | 25% | 58% | High |
| Klarna KLARCut about 40 percent of its workforce citing AI and quantified a roughly 40 million dollar profit gain, then partly reversed the deepest customer-service cuts on quality grounds, with AI-attributable new revenue coming only indirectly through existing BNPL fees. | FintechSE | The Efficiency TrapEvidence only: Reinventors | 67% | 42% | Med |
| Klaviyo KVYOStates that AI extends to how it operates and reports code shipped per engineer nearly doubling, but never says where the freed capacity went and puts no revenue figure on its Composer agent. | Marketing SaaSUS | WatchingEvidence only: Fragile Growth | 42% | 42% | Med |
| Koninklijke Philips PHIARuns cloud AI imaging across more than 1,000 North American hospitals and is shifting Informatics to recurring revenue, but attributes its 0.8bn euro 2025 productivity savings to cost management rather than AI. | Medical technologyNL | WatchingEvidence only: Fragile Growth | 17% | 42% | Med |
| Lonza Group LONNRuns digital-twin and machine-learning process work and launched the model-based Design2Optimize platform, but all of it supports standard CDMO revenue and none is quantified. | CDMOCH | Watching | 17% | 8% | Med |
| Masimo MASITalks up AI-enabled next-generation monitoring through a 2025 Philips partnership, but discloses no internal AI efficiency and no realised AI revenue line yet. | Patient monitoringUS | Watching | 8% | 25% | Low |
| Mettler-Toledo MTDRuns a live AI knowledge base that lifts service engineers' first-fix rates, but sells no AI product and quantifies no saving. | Lab instrumentsUS/CH | Watching | 17% | 0% | Med |
| Monday.com MNDYDoubled AI-related recurring revenue in a single quarter to 17 percent of net new ARR on a new seats-plus-credits model, while cutting about 20 percent of staff for roughly 100 million dollars of savings that leadership attributes to org friction rather than AI. | B2B SaaSIL | Fragile Growth | 33% | 50% | High |
| MongoDB MDBSells AI on two separate published meters, per-token Voyage embeddings and dedicated vector search nodes, and reports vector search outpacing company growth, but names no internal AI tool across four earnings calls and attributes its one restructuring to spend discipline rather than AI. | DatabaseUS | Fragile Growth | 17% | 50% | Med |
| Natera NTRASays it is deploying AI at a frenetic pace to automate its large lab and lift gross margin toward 70 percent, but attributes none of its 38 percent revenue growth or its Signatera franchise to AI. | DiagnosticsUS | WatchingEvidence only: The Efficiency Trap | 33% | 0% | Med |
| NCC Group NCCSells AI security to Meta and Google under a named Securing AI proposition, but attributes every quantified efficiency gain to Manila offshoring and the Escode separation, and the CEO's AI innovation bonus objective was scored only half-achieved. | Cyber securityUK | WatchingEvidence only: Fragile Growth | 33% | 33% | Med |
| NICE NICEGrew AI ARR 66 percent to 328 million dollars with AI in every new seven-figure CXone deal, but its own internal AI use is a single unquantified line in the 20-F. | Contact centre AIIL/US | Fragile Growth | 8% | 50% | Med |
| Notion privateReports roughly 700 internal AI agents across a 1,000-person company and folds monetised AI and credit-billed Custom Agents into its Business tier, but quantifies no savings and no AI revenue funding model. | Productivity SaaSUS | WatchingEvidence only: Reinventors | 25% | 42% | Med |
| Nuix NXLNuix Neo, its AI-enriched platform, grew contract value 148% to a fifth of the company, while efficiency-side AI gets one unquantified line. | Investigative analyticsAU | Fragile Growth | 25% | 50% | High |
| Omada Health OMDATies a record 73 percent quarterly gross margin partly to AI in its care teams yet its CEO refuses to monetise AI as standalone care, a clear Efficiency Trap. | Digital healthUS | The Efficiency Trap | 75% | 8% | Med |
| Oxford Instruments OXIGDiscloses detailed operational-excellence savings with no AI in them and treats AI only as an R&D priority to build into products, so nothing is live. | Scientific instrumentsUK | Watching | 0% | 17% | Low |
| Oxford Nanopore Technologies ONTMachine-learning basecalling underpins the entire sequencing franchise, with 2025 revenue up 24 percent and clinical up around 60 percent, yet disclosed margin gains are attributed to a business-model shift, not AI. | SequencingUK | Fragile Growth | 17% | 50% | Med |
| Palantir Technologies PLTRAIP drove 149 percent US commercial revenue growth in Q2 2026, but Palantir discloses no measured saving or freed capacity from applying AI to its own operations, so its evidenced story is almost entirely on the revenue side. | AI and data softwareUS | Fragile Growth | 17% | 50% | Med |
| Palo Alto Networks PANWPrisma AIRS is described as the fastest-scaling product in company history with Cortex XSIAM past 600 million dollars ARR, but internal efficiency is only a qualitative claim of SOC teams moving from triage to strategic defence with no headcount reduction. | Cyber securityUS | Fragile Growth | 33% | 50% | Med |
| Personio privateShips embedded AI features and bought a recruiting-AI startup, but reports no distinct AI revenue line and nothing about internal AI efficiency. | HR SaaSDE | Not enough evidence | 0% | 17% | Low |
| Phreesia PHRUses AI internally to automate manual work and cut outsourced spend, a named driver of margin expansion with a roughly $37M efficiency target, while its Voice AI revenue is live but earlier, placing it in the Efficiency Trap. | Healthcare ITUS | The Efficiency TrapEvidence only: Reinventors | 75% | 33% | Med |
| Qiagen QGENRuns five live AI-enabled software applications and ring-fences about 20 percent of its digital-insights sales to fund a roadmap to at least 14 by 2028, but discloses no AI use in its own operations. | Life science toolsNL/DE | Fragile Growth | 0% | 58% | Med |
| Recursion Pharmaceuticals RXRXRecursion runs drug discovery on a full-stack AI operating system and has booked over $500m in AI-partnership milestones, but its 2025 cost cuts were restructuring rather than disclosed AI efficiency. | AI drug discoveryUS | WatchingEvidence only: Reinventors | 33% | 33% | Med |
| RELX RELAI-enabled products are around 90% of the value of new LexisNexis sales and drove a 10% first-half revenue rise, but RELX discloses no AI-attributable saving in its own operations. | Information and analyticsUK | WatchingEvidence only: Fragile Growth | 17% | 42% | Med |
| Renishaw RSWNames AI-driven chip demand as a growth driver and flags competitors' use of AI as a business risk, while its own adoption is limited to investing in AI productivity tools with no AI-enabled offering disclosed. | Precision instrumentsUK | Watching | 8% | 0% | High |
| Repligen RGENAdds AI to its bioprocessing products through the Novasign partnership and lists AI implementation in a transformation office, but everything is early and nothing is AI-specific in the numbers. | Life science toolsUS | Watching | 8% | 25% | Low |
| ResMed RMDIts Dawn generative-AI assistant cut live human contact-centre queries by 40 percent, a rare tracked efficiency result, but its AI revenue is still an embedded feature rather than a new line. | Connected medical devicesUS | Watching | 42% | 17% | Med |
| Revvity RVTYUses internal AI to speed its own software delivery and sells the new Signals AI framework, but frames the cost benefit as years out while its Signals software revenue fell about 20 percent. | Life science toolsUS | WatchingEvidence only: Fragile Growth | 25% | 33% | Med |
| Sage Group SGENames AI across four internal functions and reports engineers saving over 160,000 hours in six months, but AI revenue stays early with no discrete line disclosed. | Business softwareUK | The Efficiency Trap | 67% | 25% | Med |
| Salesforce CRMAgentforce passed 1 billion dollars in run rate while AI handling of half its support interactions let Benioff cut support headcount from about 9,000 to 5,000. | B2B SaaSUS | Reinventors | 83% | 58% | High |
| Samsara IOTSamsara's AI models run on 25 trillion data points and its emerging AI products passed $100m in ARR, but it reports no AI improvement to its own internal operations. | Connected operationsUS | Fragile Growth | 0% | 50% | Med |
| SAP SAPReports about 30 percent internal developer productivity gains and a hiring slowdown while AI features are attached to more than 90 percent of its largest deals. | Enterprise softwareDE | The Efficiency TrapEvidence only: Reinventors | 75% | 42% | Med |
| Sartorius SRTSells AI-powered Incucyte software modules today but its own AI use is a single warehouse robotics proof of concept with SAP, so it sits in Watching. | Life science toolsDE | WatchingEvidence only: Fragile Growth | 17% | 33% | Med |
| Schrodinger SDGRSells the new Bunsen AI co-scientist and other scaling AI-enabled products and uses AI in its own drug discovery, but its operating-cost cuts are attributed to expense discipline rather than to AI. | Computational drug discoveryUS | Fragile Growth | 42% | 50% | Med |
| ServiceNow NOWRuns about 450,000 of its own AI agents for a stated 100 million dollars of 2025 savings while Now Assist reached about 600 million dollars in AI contract value. | B2B SaaSUS | Reinventors | 92% | 50% | High |
| Shopify SHOPMade reflexive AI use a company-wide baseline expectation tied to performance reviews and headcount requests, but keeps its AI features free and discloses no measured efficiency outcome, so the conservative score stays in Watching. | Commerce SaaSCA | WatchingEvidence only: The Efficiency Trap | 33% | 25% | Med |
| Siemens Healthineers SHLBuilt its new strategy phase around healthcare AI with 6 to 9 percent mid-term growth targets and sells the AI-Rad Companion family via teamplay, but discloses no AI-specific internal savings. | Medical technologyDE | WatchingEvidence only: Fragile Growth | 17% | 42% | Med |
| Simulations Plus SLPSells AI-driven drug-discovery software and has ring-fenced NIH grant funding for its AIDD module, but discloses no AI use in its own operations, so coverage is low. | Biosimulation softwareUS | Not enough evidence | 0% | 58% | Med |
| Smartsheet privatePrivate since the January 2025 Blackstone and Vista buyout, Smartsheet has shipped a full agentic AI suite but published nothing on internal AI use, savings or AI monetisation. | B2B SaaSUS | Not enough evidence | 0% | 17% | Low |
| Snowflake SNOWNames measured internal AI gains across seven functions including support, engineering, finance and HR, took a roughly 200-person reduction in force and then hired 17 people organically in a quarter, and its CFO says the AI assistant was the largest driver of a raised full-year guide. | Data cloudUS | Reinventors | 83% | 58% | Med |
| Softcat SCTRides AI datacentre demand to 79% hardware growth and bought its way into AI consulting with Oakland, but its only quantified internal savings sit in a Microsoft marketing case study and headcount is growing in every function. | IT resellerUK | WatchingEvidence only: Fragile Growth | 25% | 42% | Med |
| Spectris SXSShips AI inside the Mastersizer 3000+ and sells AI/ML-enabled engineering solutions through HBK, but its internal programme is a data platform described only as a foundation for AI rather than deployed AI. | Precision instrumentationUK | WatchingEvidence only: Fragile Growth | 8% | 33% | High |
| Tecan TECNShips AI product from Veya to the NVIDIA-powered Introspect platform but attributes none of its operational-excellence programme to AI. | Lab automationCH | WatchingEvidence only: Fragile Growth | 0% | 33% | Med |
| Teladoc Health TDOCUses AI-assisted documentation to cut therapist admin time at BetterHelp, but its AI-driven new products were still launching and no AI-attributable revenue or savings figure is disclosed. | Digital healthUS | Watching | 33% | 17% | Med |
| Temenos TEMNDocuments internal AI in its development lifecycle, support workflows and firm-wide tooling and ships AI agents and copilots across its product line, but attaches no number to any of it. | Banking softwareCH | WatchingEvidence only: Fragile Growth | 25% | 33% | Med |
| Tempus AI TEMGrew AI-native Data and Applications revenue 28 percent to $93M and signed about $200M in new AI data-and-modelling licences with pharma including BioNTech, but discloses no use of AI to run itself. | Precision medicine dataUS | Fragile Growth | 0% | 58% | Med |
| The Trade Desk TTDKokai, The Trade Desk's AI-powered platform, is used by about 85% of clients, but the revenue runs through the same take rate and the company discloses no internal AI efficiency. | Ad techUS | WatchingEvidence only: Fragile Growth | 0% | 33% | Med |
| Thermo Fisher Scientific TMOEmbeds OpenAI across product development, service and operations and sells AI drug-development tools such as OSDPredict, but discloses no realised savings and no AI-specific revenue funding model. | Life science toolsUS | WatchingEvidence only: Fragile Growth | 25% | 33% | Med |
| Twilio TWLOThe CEO reports a 15% employee productivity gain from internal AI across coding, support and inbound sales, while the formal earnings disclosure credits efficiency to carrier deals and a cloud migration and puts no number on AI revenue. | Communications APIUS | The Efficiency TrapEvidence only: Reinventors | 50% | 33% | Med |
| UiPath PATHRuns its own agents across engineering, support and internal operations and sells a consumption-priced agentic platform, but leaves freed capacity and its AI funding model undisclosed, so the harsh headline reads Watching while the evidenced picture is Reinventors. | AutomationUS/RO | WatchingEvidence only: Reinventors | 33% | 42% | High |
| Unity Software URuns its entire growth story through an AI ad platform that passed a $1 billion run rate two quarters early and a consumption-priced AI assistant, while disclosing nothing about AI in its own operations. | Real-time 3DUS | Fragile Growth | 0% | 50% | Med |
| Veeva Systems VEEVIs remaking its product line around AI, with the Ostro acquisition and usage-priced agents live, but discloses no use of AI in its own operations and grew headcount in the quarter. | Life sciences SaaSUS | Fragile Growth | 0% | 50% | Med |
| Veradigm MDRXUses AI inside its own revenue-cycle teams and to scale real-world evidence while selling AI products like Ambient Scribe, scoring an Efficiency Trap that renormalises to Reinventors on evidence only. | Healthcare ITUS | The Efficiency TrapEvidence only: Reinventors | 50% | 42% | Med |
| Waters Corporation WATDescribes AI only as being tested for QC, troubleshooting and customer support, with no realised savings and no AI revenue line, so it sits firmly in Watching. | Life science toolsUS | Watching | 17% | 8% | Low |
| Wise WISEUses AI internally across fraud, KYC and customer support through a multi-provider LLM gateway, but sells payments rather than AI and reports no AI-attributable new revenue, the only Efficiency Trap in the batch. | FintechUK | WatchingEvidence only: The Efficiency Trap | 25% | 0% | Med |
| WiseTech Global WTCPlans to cut roughly 2,000 roles as it becomes an AI-led organisation and has moved 95% of customers to transaction pricing with embedded AI features, but attributes no revenue to AI specifically. | Logistics softwareAU/US | The Efficiency TrapEvidence only: Reinventors | 58% | 42% | High |
| Workday WDAYReports agentic AI new ACV up more than 200 percent year over year with over 4,000 customers using an agent, but discloses little about using AI to cut its own costs. | B2B SaaSUS | WatchingEvidence only: Fragile Growth | 17% | 42% | Med |
| Xero XRORevenue per FTE rose 21 percent on flat organic headcount with emerging internal AI efficiencies, but Xero keeps its JAX agent free to subscribers so no discrete AI revenue is disclosed. | Accounting SaaSNZ/AU | Watching | 25% | 17% | Med |
| Xplor Technologies privateIs building AI-native payroll and reporting tooling after acquiring Bitlancer, but the AI revenue is still in build and internal use is undisclosed. | Payments and SaaSUS | Not enough evidence | 0% | 17% | Low |
| Zoom Communications ZMThe CFO calls Zoom its own customer zero and says AI removed costs from its own support organisation, yet AI appears in no published margin bridge and the CFO declines on the record to give an AI revenue figure. | Communications SaaSUS | Fragile Growth | 25% | 50% | Med |
| Zscaler ZSZscaler's AI Security ARR passed its $400m target three quarters early and is heading past $500m, but the company discloses no internal AI efficiency programme and no realised operational savings. | Cyber securityUS | Fragile Growth | 0% | 50% | Med |
No companies match. Show all
// Method
How the scores were made.
- Six questions, the self-assessment's scales. Three on efficiency (deployment, savings against expectation, freed capacity) and three on new revenue (whether AI made new revenue possible, what kinds, and how new bets are funded). Each axis is scored out of 12 and shown as a percentage.
- A status on every answer. Evidenced, absent or undisclosed. Undisclosed is not the same as zero, which is why two scores are kept: silence counted as zero, and silence set aside.
- A coverage gate. A company only counts in the quadrant numbers if the two questions that decide its quadrant, efficiency deployment and new revenue, are evidenced or absent. 12 companies fail that gate and are shown as not enough evidence.
- Sources on every score. Annual reports and 10-K or 20-F filings first, then earnings calls and investor presentations, company announcements and product pages, named-executive interviews and specific trade press. No analyst speculation and no supplier-written case studies. 533 source links in all. Confidence is set by source quality: 22 high, 81 medium, 17 low.
- A known bias. This measures what a company discloses, not what it does. Large listed companies disclose more than private ones, and efficiency work is disclosed more readily than revenue experiments that failed. Both push the numbers in a direction you can name.
This research is part of the field guide AI and the Board: the questions a board should ask about AI, and the evidence behind them.
Read the guide// See also