When a posting expects AI use, are security managers asked to be more hands-on?Manager postings that expect AI ask for hands-on work more often.
Written entirely by agents in the Research Factory, within guardrails and controls set by our team. Figures as of 1 October 2026.
Read the full research paperMore often, at the manager level. In fall 2026, 78% of 32 security manager postings that expected AI use asked the hire to stay hands-on. Among 47 manager postings with no AI expectation, 43% did.
At the executive level, a difference is too close to call. Executive postings that expect AI make deciding the core of the job in 81% of 32 cases, against 65% of 37 without, but the range of likely values includes zero.
These figures come from a fixed panel of 149 security people-manager postings at 120 companies. The postings cannot show what produces a difference. AI-exposed postings come from different teams and companies.
These findings come from one panel of 149 security people-manager postings at 120 companies, mostly in security operations and GRC, and 90 of them at enterprises.
The answer at a glance
Manager-level postings that call out AI ask the hire to be hands-on more often than those that don't: 78% of the 32 that call out AI, against 43% of the 47 that don't.
About half of security people-manager postings ask the hire to stay hands-on: 49% of 149.
2026 manager postings that expect AI ask for hands-on work more often than 2024 manager postings: 78% against 52%.
Executive postings may center on making decisions more often when they expect AI: 81% of those that do, against 65% of those that don't. The gap is too close to call.
A gap in coordination work between manager postings with and without AI, 25% against 40%, is also too close to call.
The numbers
Three AI models each read all 149 postings. The answer used is the one at least two of the three agree on.
| Level | AI expectation in the posting | Postings | Stays hands-on | Core is deciding | Core is coordinating |
|---|---|---|---|---|---|
| Manager | No | 47 | 43% | 30% | 40% |
| Manager | Yes | 32 | 78% | 19% | 25% |
| Executive | No | 37 | 41% | 65% | 11% |
| Executive | Yes | 32 | 38% | 81% | 3% |
| All people managers | 149 | 49% | 47% | 21% |
Source: AKA Security analysis of 149 security people-manager postings at 120 companies in AKA's job-board data (open on or after 23 September 2026), read by three AI models.
An AI expectation means the posting asks the hire to use AI tools, build AI automation, secure AI systems, govern AI or bring machine-learning skills. Stays hands-on means the posting says the hire mainly does the technical work personally, such as a player-coach who works real cases, builds detections or does the compliance work. The core is the activity the responsibilities weight most heavily.
Manager postings that expect AI ask for hands-on work
In security manager postings open in fall 2026, 78% of the 32 that expected AI asked the hire to stay hands-on. Of the 47 with no AI expectation, 43% did. The gap is +36 points, with a range of likely values of +13 to +56 points. The 32 postings come from 28 companies and the 47 from 35.
The gap holds for each AI model alone, at +24, +33 and +46 points. It also holds with AI-native companies, AI infrastructure companies and security vendors removed, at +37 points.
The finding is rated Medium. In a split-model check, one model judged the AI expectation and another judged the work mix, so no single model judged both sides. That gap is +22 points, with a range of −1 to +43 that includes zero. The three models also differ on size by up to 14 points.
Across all 149 people-manager postings, 49% ask the hire to stay hands-on. The range is 40% to 58%, and each model alone puts the share between 49% and 51%.
Against spring 2024
The 2024 comparison uses 72 people-manager postings from a public US job-postings dataset, listed 5 to 19 April 2024. Of those, 44 are at manager level, from 40 companies. AI expectations appear in 3% of the 72.
In 2024, 52% of US security manager postings asked the hire to stay hands-on. In 2026, 78% of AI-exposed manager postings did. The difference is +26 points, with a range of +4 to +46.
| Comparison | 2024 | 2026 | Range of likely values for the change, points |
|---|---|---|---|
| Stays hands-on: all 2024 managers, 2026 managers with an AI expectation | 52% | 78% | +4 to +46 |
| Stays hands-on: all 2024 managers, 2026 managers with no AI expectation | 52% | 43% | −32 to +13 |
| Coordinating: all 2024 managers, 2026 managers with no AI expectation | 41% | 40% | −20 to +20 |
| Coordinating: all 2024 managers, 2026 managers with an AI expectation | 41% | 25% | −36 to +7 |
| Executive, AI expectation, deciding | 68% | 81% | −9 to +36 |
| Manager, deciding, all postings | 11% | 25% | +1 to +26 |
| Each bar is a range of likely values against zero, the dashed line. A hollow bar includes zero: we can’t tell the change from no change. | |||
Source: AKA Security analysis of 72 US people-manager postings from a public job-postings dataset (April 2024) and 149 people-manager postings in AKA's job-board data (fall 2026).
Manager postings without an AI expectation show no clear change from 2024. Deciding as the core of manager postings rose from 11% of 44 in 2024 to 25% of 79 in 2026, a likely but borderline rise.
Executive postings that expect AI: A gap in decisions is too close to call
In security executive postings open in fall 2026, 81% of the 32 that expected AI made deciding the core of the job. Of the 37 with no AI expectation, 65% did. The gap, +16 points, has a range of −4 to +36, so it is too close to call.
The models split on this one. Two of the three models alone find a clear gap, and one does not. Agreement on the deciding label runs from 0.49 to 0.69 across the model pairs. Coordinating was the core of 3% of executive postings with AI and 11% without.
The coordination gap is too close to call
Among security manager postings, coordinating was the core in 25% of the 32 that expected AI and 40% of the 47 without. The gap, −15 points, has a range of −36 to +6.
Agreement on the coordinating label is low. One hosted model is the outlier, and no model pair that includes it agrees above a kappa of 0.39. Deciding as the core of manager postings went from 30% without AI to 19% with AI, which is also too close to call.
A second read, two days later
On 25 September 2026, two hosted models read the US-headquartered postings again with newer settings. Only postings where both agree the hire manages people, and agree on the AI expectation, count.
| Second read, US postings | Postings, no AI and AI | Hosted model A | Hosted model B |
|---|---|---|---|
| Manager hands-on, AI minus no AI | 32 and 24 | +42 (+16 to +64) | +41 (+15 to +64) |
| Executive deciding, AI minus no AI | 20 and 22 | +11 (−9 to +33) | +25 (+10 to +43) |
| 2026 AI-exposed managers against 2024 managers, hands-on | 44 (2024) and 24 | +20 (−2 to +40) | +27 (+3 to +49) |
Source: AKA Security analysis of US-headquartered people-manager postings in AKA's job-board data (read 25 September 2026) and 44 US manager postings from a public job-postings dataset (April 2024). Figures are gaps in points, with the range of likely values.
The manager hands-on gap holds in both second reads. The executive decision gap and the 2024 comparison hold in one and not the other.
The question as asked
With AI, there’s a general hypothesis that managers’ roles will change to do less coordination and more decision-making and in some cases more doing. Does the data show that? More broadly, how is the role of people managers in security changing across middle-management and executive management levels?
Asked by Lenny Zeltser
Where the data comes from
The 2026 panel is 149 security people-manager postings at 120 companies, 90 of them at enterprises. The postings were open on or after 23 September 2026, and the panel does not refresh with daily data. By level, 79 are manager postings and 69 executive postings, with 1 at another level. By function, 53 are in security operations, 51 in GRC, 21 in privacy and 11 in offensive security.
The panel was built over 54 days. 124 of the 149 postings were first seen after the collection widened to more titles on the evening of 21 September 2026.
It started from 585 postings with a leadership title: manager, director, head, chief, VP or lead, after removing program, product, project, account and risk managers. Of those, 166 were security people managers with reports. A local model ruled 14 of them not security roles, and 3 more were removed as job-aggregator postings or postings closed before 23 September.
The 2024 set starts from 245 leadership-titled US postings. A scope check ruled 100 of them not security roles, such as financial audit, model risk and physical security, leaving 72 people managers after aggregators were removed.
Three hosted AI models each read all 149 postings. Agreement across the model pairs is 0.74 to 0.84 on hands-on work, 0.86 to 0.92 on AI expectation, 0.49 to 0.69 on deciding and 0.39 to 0.78 on coordinating. Each range of likely values (95%) is found by redrawing the sample 2,000 times, a company at a time.
How this answer was stress-tested
Main results, AI minus no AI, in points: manager hands-on +36, executive deciding +16, manager coordinating −15.
| Test | What it found |
|---|---|
| Each model alone, manager hands-on. | +24 (+1 to +45), +33 (+12 to +53) and +46 (+26 to +64). Same direction, each clear of zero. |
| Split-model check, manager hands-on. The AI label from one model, the work-mix label from another. | +22 (−1 to +43). No clear change, smaller by 14 points. |
| Without AI-native companies, AI infrastructure companies and security vendors, manager hands-on. | +37 (+8 to +63). Same direction, about the same size. |
| Each model alone, executive deciding. | +30 (+10 to +50) and +24 (+5 to +42) are clear. +16 (−5 to +36) is not. |
| Split-model check, executive deciding. | +13 (−7 to +34). No clear change. |
| Without AI-native companies, AI infrastructure companies and security vendors, executive deciding. | +23 (+1 to +44). Likely but borderline. |
| Second read, 25 September. | Manager hands-on +42 and +41, both clear. Executive deciding +11, not clear, and +25, clear. |
| Second read against 2024. | +20 (−2 to +40), not clear, and +27 (+3 to +49), clear. |
| Stricter title rule. Titles naming a domain called "management" not counted as managers. | The panel moved by 1 point at most. |
| Scope check of the 2024 set. | 100 of 245 postings were not security roles. Hands-on agreement is lower in 2024: 0.47 to 0.64, against 0.74 to 0.84 in 2026. |
| Scope check of the 2026 panel. A local model read the 166 postings. | 14 ruled out and 3 more removed, leaving 149. |
What would strengthen this answer
| Test | What it would settle | Why it has not run |
|---|---|---|
| A larger manager sample | Narrower ranges, and a split by function beyond operations and GRC | |
| The same teams over time | Whether a team's manager posting became more hands-on after the team added AI expectations | |
| Matching on company and function | Separates AI from the kind of team that tends to post AI expectations | |
| What managers actually do | Whether stated expectations match the job as worked | |
| A practitioner read of 50 postings | Whether "stays hands-on" matches how a security leader reads the role |
Other formats
The same findings and figures, laid out as a research paper with numbered sections, references and a citation.
Read the full research paperVersion history
Every change to this piece is tagged and logged. When a figure moves or a finding no longer holds, it is recorded here rather than edited in silence.
- The base of the 78% and 43% comparison is now named: 32 and 47 postings among 79 managers.
- Manager postings that expect AI ask for hands-on work in 78% of cases, against 43% without. The early-access draft and its title said 76%. The collector widened on the evening of 21 September, and 124 postings were first seen after the widening, where the early-access draft said 129.
- When a posting expects AI use, are security managers asked to be more hands-on? Panel figures as of 23 September 2026, second read 25 September.
Corrections
Received. The Research Factory rechecks the figure against the source, and any fix appears in the changelog.
The rating scales and the standard checks are the same for every piece. How the research is made.
Ask the Research Factory a question.
This is not a live chat. Accepted questions become new pieces, published in a later release.
- 01 You ask, with the decision it would inform You queued
- 02 We check job postings can answer it Our team queued
- 03 Approved questions join the reader queue Our team queued
- 04 Agents write the analysis, ratings and piece Research Factory queued
- 05 Every answer runs our standard controls Our controls queued
- 06 Published and tagged, credited if you want Research Factory queued
- 01 You ask, with the decision it would inform You done
- 02 We check job postings can answer it Our team next
- 03 Approved questions join the reader queue Our team queued
- 04 Agents write the analysis, ratings and piece Research Factory queued
- 05 Every answer runs our standard controls Our controls queued
- 06 Published and tagged, credited if you want Research Factory queued