Translate the work; do not abandon it.
Established language still holds most of the attention. AI-framed terms provide a new route into capabilities many risk, security and resilience teams already deliver.
Eudai · Research explorer · September 2026
The work of risk is not disappearing. It is being embedded into how every function now runs.
An interactive companion to the report. Explore eight lenses on the same market—search demand, regulation, certifications, analyst coverage, disclosure language, employment, capital and public procurement—each moving at its own pace and revealing a different signal.
The headline finding
AI-framed language grew 302% and established language fell 13%. But comparable business vocabulary fell more than twice as far.
Growth in AI-framed search language since 2023. Over the past year, established vocabulary declined 29%.
Established language still receives 8.3 times more monthly search attention across the six concept families.
AI-framed language is gaining ground in risk and security roughly twice as fast as in comparable business functions.
The argument
Risk, security and resilience have always been loops: govern, detect, respond, recover, learn, assure, and round again. That is what a management system is.
Product, operations, strategy, software delivery, marketing, support and AI development are converging on the same operating model: define the goals that must be protected, detect when they slip, respond, correct, retain the lesson and prove the cycle worked. The jobs differ, but the stages do not.
Select a stage to see the vocabulary underneath it.
Stage 1 of 6
The stage growing fastest in absolute attention, and the one the AI vocabulary entered first. A management system is a loop by definition, which is why ISO 42001 arriving matters more than its certificate count yet shows.
Language built around documents, events and job titles: things created, held and then filed away.
| crisis management | 20,958 | −27% |
| enterprise risk management | 14,842 | −54% |
| incident management | 14,092 | −34% |
| security operations center | 13,500 | −13% |
| business continuity manager | 10,292 | −88% |
| disaster recovery plan | 10,133 | −30% |
| incident response plan | 6,092 | −19% |
| audit management software | 1,582 | −23% |
Language built around services, postures, monitored states and management systems: things that operate continuously.
| ai governance | 25,442 | +90% |
| nist ai rmf | 10,200 | +190% |
| jailbreak detection | 8,967 | +611% |
| ai readiness | 7,133 | +120% |
| security operations | 5,117 | +27% |
| ai observability | 4,150 | +266% |
| ai agent security | 3,758 | +529% |
| ai monitoring | 3,167 | +37% |
Language forms as the market works out what the problem is and who owns it.
| human in the loop | 53,175 | +623% |
| team learning | 8,200 | +73% |
| human on the loop | 2,075 | +399% |
Once something is taken for granted, people stop searching for it.
| lessons learned | 66,125 | −26% |
| feedback loop | 36,667 | −2% |
| continuous improvement | 36,417 | −45% |
| after action review | 5,267 | −27% |
| continuous auditing | 1,157 | −63% |
Traditional language is losing searches across business functions, not just risk and security. AI language is growing about twice as fast here. The same learning loop is appearing everywhere.
Business continuity appears frequently in company disclosures but rarely as a certified capability. The work is widely expected, but formal investment in it is much harder to see.
10-K results containing “business continuity” in 2024 alone: one phrase, one year and US filers only. Continuity risk is routinely acknowledged in formal company disclosures.
Valid ISO 22301 certificates worldwide in 2024: the total certified population across all countries.
These figures measure different things, but the contrast still matters. Filings are documents produced during a single year in one jurisdiction. Certificates are the accumulated global total. They are not directly comparable, but they show how widely continuity risk is acknowledged and how rarely the capability is formally certified.
Disclosure language is changing too, although more slowly because filings are often lawyer-drafted and carried forward from year to year. Language describing a document grew 19%. Language describing capabilities that must operate continuously grew roughly ten times faster.
2023 to 2026 YTD. SEC EDGAR exact-phrase 10-K result documents — documents, not issuers, and 2026 is a partial year. ISO Survey of Management System Standard Certifications 2020–2024.
Risk, resilience and assurance become part of how every function operates. The work spreads across the organization, while the traditional departments and titles become less visible. Risk becomes a property of the work rather than a separate review of it. This is the direction suggested by the search data.
Risk, security and resilience professionals take responsibility for making these loops dependable. They help define what must remain true, detect when it changes and ensure the organization can respond, recover and learn. That is a larger role than owning a review gate, and one no function clearly owns today.
Titles are being added. A job-posting analysis covering August 2022 to August 2026 shows AI-governance roles being created quickly. At the level of job titles, AI governance is adding work rather than replacing it.
Responsibilities are converging. Thirteen of the 43 postings — roughly three in ten — place model risk, operational resilience or incident response inside AI-titled positions. AI governance is being built into existing risk and resilience structures, not developing entirely on its own.
The absence is just as revealing. No posting in the corpus uses “continuous assurance” or “organizational learning” as a job title. “Lessons learned” appears as a responsibility inside a resilience role, but never as the name of one. The loop is visible in the work, not in the title—the same pattern found in the search data.
One term remains unexplained. “BC DR” attracts 278,042 monthly searches and grew 37%, accounting for the continuity category’s entire gain. Because it abbreviates two episodic disciplines, the learning-loop thesis does not explain it — and nor does abbreviation on its own.
| Acronym | Change | Spelled out |
|---|---|---|
| SOAR | flat | — |
| CISO | −6% | — |
| DFIR | flat | −16% |
| CSIRT | flat | −5% |
| SIEM | −31% | −31% |
| PSIM | +12% | −8% |
| BC DR | +37% | −30% |
The difference may be audience. Acronyms are generally used by people already inside a discipline; spelled-out terms are more likely to help outsiders discover it. Insider language may be holding while the wider discovery path moves elsewhere. But BC DR is also a short, potentially ambiguous query, so the report treats it as an unresolved exception.
What leadership should take from this
8 sources track the same market at different speeds. The strongest conclusions appear where several of them point in the same direction.
Established language still holds most of the attention. AI-framed terms provide a new route into capabilities many risk, security and resilience teams already deliver.
Regulatory interest builds well before the obligation takes effect. DORA-related attention has fallen to 24% of its peak, while regimes with deadlines still ahead remain at 82% and are rising.
Assessment, audit and certification language grew across both emerging and established subjects. Platform, software and tools language grew only when attached to the right subject.
Search attention for ISO 42001 overtook ISO 22301 in September 2025. Comparable official certification data has yet to show how much adoption followed.
Business continuity still dominates company filings. But third-party risk management, cyber incident response and operational resilience are gaining ground fastest.
Demand is broadly consistent throughout the year. Timing matters less than relevance.
Evidence explorer
Explore each source in the unit it actually measures. Where assumptions affect a result, adjust them and see what changes. The explorer keeps attention, adoption, analyst coverage and disclosure separate rather than forcing them into a single market score.
Search attention
Google Keyword Planner measures how people search, not what they buy, spend or adopt.
Reported cluster endpoints, indexed to 2023 = 100. No unreported monthly values have been interpolated.
Source: Resilience 2030, Figure 1; Google Keyword Planner, 1,621 terms, Sep 2022–Aug 2026.
The translation
The search volumes have not crossed, but the language is already changing. These paired terms show different ways of finding related work; they do not prove that one term replaced the other.
Priority wedges
A way to identify promising areas for further investigation—not a measure of market size or a recommendation to buy.
| Why it matters |
|---|
The work of risk is not disappearing.
It is being embedded into how every function now runs.
Every function that now runs a loop performs governance, detection, response, recovery and assurance under its own vocabulary.
An SRE setting an error budget is performing risk appetite.
A leadership team resetting OKRs each quarter is performing governance review.
A growth team running a holdout is performing control testing.
An AI team running evals is performing assurance.
None of them call it that, and none of them report to a risk function. The phrases that name the loop as an idea are falling — continuous improvement −45%, lessons learned −26% — while “human in the loop” grows 623%.
What is contested is no longer whether the loop exists, but the job inside it and who owns it.
Three myths
Not across the market. The all-term search index is flat. Urgency, threats and regulatory timing shape the exceptions.
Search measures questions. Certificates, analyst coverage and filings show different layers, and none alone equates to revenue.
AI vocabulary is growing fast from a smaller base. Much of it renames familiar governance, assurance and resilience work.
Method
Each source keeps its own unit. Search measures monthly attention; certification, accumulated certificates; regulation, timing against peak attention; analyst coverage, position within each firm’s taxonomy; filings, matching documents; employment, patterns within 43 sourced postings plus market aggregates; capital, 449 source-linked transactions; and public procurement, 737 government awards. The employment data is not a census or quarterly demand series.
The control. We compared 125 terms from recruiting, accounting, marketing, support and software delivery using identical settings. AI-framed language grew faster in risk and security, while established language declined across all functions. Medians are used because “AI chatbot” dominates the control group’s total volume.
Minimum base. Keyword Planner reports rounded values, making growth unreliable at very low volumes. Percentages are shown only where the prior base was at least 90 monthly searches and the history contained enough variation. Smaller terms are identified without a growth rate.
Google Keyword Planner. 1,621 terms, monthly, September 2022 to August 2026, brand names excluded from cluster totals.
The ISO Survey of Management System Standard Certifications, 2020–2024 (ISO/CASCO, from IAF CertSearch).
Published compliance dates for DORA, NIS2, the EU AI Act and the Cyber Resilience Act.
Gartner and Forrester report metadata, 2010–2026, public sources only — 176 report rows.
SEC EDGAR full-text search. 10-K results for 8 exact phrases, August 2022 to July 2026 — 13,675 documents.
What to do about it
When the language around your category changes, the first challenge is positioning and discovery—not rebuilding the product. That is where we work.
This explorer makes the measurement boundary visible rather than papering over it. No chart should be read as a market-size estimate, and none proves a causal relationship.