Editorial note: This is a composite narrative. “Sandra” is not a specific real individual. Her story is a composite, drawn from career patterns Olivia has observed across many professionals navigating AI-era organizational change. Institutional affiliations and identifying details have been fictionalised. Any resemblance to a specific person or organization is unintentional.

The stories that tell me the most are never the ones people lead with.

Someone I’ll call Sandra sent me a note in April — not asking for advice, but to say thank you. Which immediately made me pay attention. She had just received a promotion she had not applied for, a title she had not expected, and an expanded scope that included work she had invented rather than inherited. She was writing because she had been reading Jackson’s Workflow Proof Playbook when she did it. And because — here is the part that stopped me — she had done it during the same quarter her company cited AI efficiency as the reason to cut 22% of its workforce. Including three of the five people on her team.

She wanted to know if that was normal. Whether there was a name for what had just happened to her. And whether she should feel guilty about it.

I told her I wanted to write about it instead. — Olivia Bennett

The Month Everything Compressed

#

Sandra had worked in marketing operations for twelve years — ten of them in B2B software. She knew how to run a campaign, manage a vendor relationship, untangle a Salesforce configuration, and produce a quarterly marketing report that actually got read. She was not the most technically skilled person in any room she had ever worked in. She was often the person who understood best what the technical output was supposed to do.

In February 2026, her company announced a 22% headcount reduction. The language in the all-hands was familiar by then: “AI-enabled efficiency gains,” “right-sizing for the next phase of growth,” “repositioning the team for greater impact.” Marketing took a proportional cut. Three of her five junior marketing coordinators were let go in the same week. One of them had started eight months earlier, fresh out of university. She had been informally mentoring him.

Sandra kept her job. And within days, the work of people whose positions no longer existed had redistributed to her desk.

What happened next is a version of the situation Raj describes in the July 3 Workplace Clinic on the AI Survivor Penalty — the expanded scope absorbed without a corresponding change in title, compensation, or staffing support. The AI tools were offered as the justification for why the math would still work. They did not explain themselves. They rarely do.

Sandra is not a single case. She is a pattern. What makes her story different is what she decided to do with the two weeks after the restructuring landed.

The Workflow She Chose

#

In week two, she sat down and listed every recurring process that now belonged to her. It was a long list. Campaign trafficking setup. Vendor communications. Monthly attribution reports. Performance decks for the commercial leadership team. And the one she kept returning to: the weekly channel performance reporting pack.

The channel pack had been Sarah’s job. Sarah, who now had no job.

Every week, it required pulling exports from three separate platforms — Salesforce, HubSpot, and a business intelligence tool — reconciling the terminology manually (because each platform categorized “lead source” differently), cleaning the resulting discrepancies, and formatting the output into a deck template the VP of Marketing could read before the Monday morning leadership sync. In a normal week, it took 3.5 hours. In weeks with data anomalies — which happened roughly once a month — it took closer to five.

Sandra hated this process with the specific animosity she reserved for things that were both painful and pointless. The painful part was the reconciliation: tedious, error-prone, and completely dependent on knowledge that only she held. The pointless part was that she had known about the terminology mismatch between the platforms for more than two years and had never had time to actually document it.

She picked the channel reporting pack.

Not because it was strategically impressive. Not because automating it would win arguments. Because it happened every week, had a visible downstream consequence, and made her miserable in a way that had a specific, addressable cause. Jackson’s playbook — which she had read twice in the weeks following the restructuring, looking for something practical to hold onto — says to pick a boring bottleneck with a visible cost. She knew exactly what she was choosing.

Two Weeks, a Dictionary, and Three Dry Runs

#

She spent the first three days of the pilot doing something that had nothing to do with AI. She built a data dictionary — a reference document mapping each field name in Salesforce to its equivalent in HubSpot to its equivalent in the BI tool. She did it by hand, field by field, because the mismatch between how each platform categorized “lead source” was the specific source of the discrepancy errors that caused the five-hour weeks. She had known about the mismatch for two years. She had never written it down.

With the dictionary complete, she used AI to build a semi-automated reconciliation template: a prompt-driven process that could ingest the raw platform exports, cross-reference them against the dictionary, flag discrepancies for human review rather than silently propagating them, and format the clean output into the deck template. She tested it against three prior weeks of historical data.

The first run produced two errors. The second run produced one. The third run matched her manual process at 98% accuracy — and took her 40 minutes instead of 3.5 hours.

She had not replaced her judgment. She had removed the parts of the process that consumed her time and produced nothing of value — the mechanical reconciliation, the formatting, the cleanup — so that she could spend more time on the part that required her: the narrative section at the top of the deck, which explained what the numbers meant to the VP of Marketing, who was not a data person and needed the week’s signals interpreted, not just displayed.

The output got better. Not because the tool improved it. Because she had time to think.

The Proof Memo

#

She had absorbed enough of Jackson’s work to know that the pilot was not the point. The documentation was.

She wrote a one-page memo with four sections.

Before: Average process duration (3.5 hours), error rate (one significant discrepancy per month causing a delay to the Monday sync), and risk profile (100% single-owner dependency — the process could not be delegated because it required reconciling knowledge that only she held).

After: Average process duration (40 minutes), error rate (zero in the three-week test period), and a new distribution structure — the cleaned output could now be reviewed by a second team member before she wrote the narrative, reducing single-point-of-failure risk for the first time since the process existed.

What I did with the time: Three things, named specifically. Reactivated a vendor contract audit that had been paused for two quarters. Began rebuilding the campaign brief template, which had not been updated since the previous marketing director left. Started attending the RevOps weekly sync she had been skipping due to time constraints — and in week three, surfaced a Salesforce field duplication issue that had been inflating pipeline attribution by approximately 8%.

What would scale: The same data dictionary approach applied to monthly attribution reporting, which had an identical platform mismatch problem. Scoped, bounded, ready to pilot.

She sent this memo to her manager in March.

Three weeks later, she was promoted to Senior Marketing Operations Manager. Her new scope: lead AI workflow optimization across the marketing organization.

Who the Practitioner Premium Actually Goes To

#

Sandra’s story is not a success story about AI. It is a success story about institutional knowledge meeting one well-chosen workflow at the right moment.

The data says something specific about this. Indeed Hiring Lab’s May 2026 snapshot found that AI-related job postings represent 5.7% of all job postings on Indeed — nearly double the previous peak. But that demand is not concentrated in a new category called “AI jobs.” It is distributed across existing functions: marketing, finance, operations, healthcare, customer success. For most organizations — the two-thirds that are still in the experimenting or piloting phase rather than at scale — the AI hire they actually need is a practitioner in an existing domain who has enough AI fluency to redesign the workflow they already own (Indeed Hiring Lab, June 18, 2026).

McKinsey’s State of AI 2025 survey confirms the mechanism on the organizational side. Companies seeing meaningful enterprise-level EBIT impact from AI are nearly three times as likely as others to have fundamentally redesigned individual workflows — not simply deployed tools onto unchanged processes. That intentional workflow redesign is listed as one of the top factors distinguishing high performers from the rest (McKinsey, November 5, 2025).

Sandra redesigned one workflow. She documented what changed. She surfaced the downstream implications — the 8% attribution inflation, the reactivated audit, the rebuilt template — of having reclaimed the time. That combination lands as something categorically different from “I use AI at work.” It is evidence of judgment, not just adoption.

The counterintuitive truth in all of this is about who has the leverage to execute that combination. It is not, as the prevailing narrative suggests, primarily the youngest person in the room with the most tool exposure. It is the person who knows why the data discrepancy exists in the first place, and what the VP of Marketing actually needs the number to say.

Sandra’s twelve years of marketing operations context was not a liability in this moment. It was the entire premise of the value she produced. No AI model trained on generic data could have built that data dictionary in three days, because no AI model knew that the legacy Salesforce configuration from two acquisitions ago still categorized paid search as “partner channel” in specific campaign types, and that this field name had never been reconciled with the HubSpot taxonomy that replaced the previous attribution model. Sandra knew this. She had been working around it for years without time to document it.

The tool needed her knowledge. Not the other way around.

The Part Nobody Warns You About

#

Sandra’s note in April had one question she kept returning to, buried toward the end: Is it wrong that I benefited from the same thing that displaced my colleagues?

I want to answer this honestly, because I think more people are sitting with a version of it than are saying it out loud.

The layoffs at Sandra’s company were framed as AI efficiency. Her promotion was framed as AI capability. Both framings use the same word. Both happened in the same quarter. That collision is not coincidental — it is structural. In most organizations right now, AI is simultaneously serving as justification for headcount reduction and as the credential for scope expansion. The people who are let go are described as having been displaced by AI efficiency. The people who are promoted are described as having demonstrated AI leadership. The organization benefits both times. The people on the different sides of that divide are left to make sense of it.

Sandra did not cause the layoffs. She also did not cause the structural reality that her company was cutting headcount while simultaneously creating a new leadership role for AI workflow optimization. Those decisions were made at levels well above her.

What she did was move through that reality deliberately rather than reactively. That is not the same as endorsing it.

There is something else worth naming about what she told me afterward. In April, the same week she received her promotion, she reached out to the junior coordinator she had been mentoring — the one let go in February — and told him what the channel reporting process had become. She walked him through the data dictionary she had built, and offered to help him document it as a portfolio piece for his job search. He had been doing that reconciliation manually for eight months; he had domain knowledge he did not know how to make legible.

The same institutional knowledge that made Sandra’s pilot work turned out to be something her laid-off colleague could point to. Made visible, it was a skill — not just a workaround. His job search got traction shortly after.

This is not a tidy resolution. But it is what the practitioner-not-specialist model actually looks like when it’s handled well: not keeping the knowledge to yourself, because that knowledge didn’t originate with you. It accumulated in a system you happened to work inside.

What the Data Doesn’t Say About Who You Are

#

When I think about who most needs to hear Sandra’s story, it is not the twenty-six-year-old who already follows every new model release and knows which benchmarks matter. That person is doing fine — or at least, they have access to the usual advice. They will figure it out.

It is the professional who has been in their domain for a decade or more and has been told, in some variation, that the AI transition is happening to them rather than for them. That their experience is a liability. That they need to catch up.

The data does not support that story. What it shows is a labor market where AI-related demand is flowing toward people who own the context — the specific, institutional, often tacit knowledge of why a process works the way it does, what the output is actually for, and where the judgment calls happen. That kind of knowledge is not replaceable by the tool. It is what makes the tool produce something credible.

Sandra at forty-three, with twelve years of marketing operations context, was better positioned for this moment than a twenty-six-year-old AI specialist who doesn’t know why the Salesforce field name doesn’t match the HubSpot field name — and doesn’t know that the mismatch has been inflating a specific metric for two years.

The advantage she built over twelve years is not what the tool displaced. In one well-chosen workflow, it was exactly what the tool required.

Have a career story you think belongs in Paths & People? The patterns I find most useful are the ones that surprised the person living them — not because things went smoothly, but because of what the outcome revealed about how the system actually works.

Email me at olivia.bennett@tlnw.uk to share your story.

References

#

• Indeed Hiring Lab. (June 18, 2026). “US Labor Market Snapshot — May 2026.” https://www.hiringlab.org/2026/06/18/us-labor-market-snapshot-may-2026/ (Accessed July 9, 2026)
• McKinsey Global Institute. (November 5, 2025). “The State of AI in 2025: Agents, Innovation, and Transformation.” https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai (Accessed July 9, 2026)
• Jackson Rodriguez. (June 29, 2026). “Career Mechanics: The Workflow Proof Playbook.” ExpertLinked. https://expertlinked.in/posts/2026-06-29-career-mechanics-workflow-proof-playbook/ (Accessed July 9, 2026)
• Olivia Bennett. (July 3, 2026). “Workplace Clinic: The AI Survivor Penalty.” ExpertLinked. https://expertlinked.in/posts/2026-07-03-workplace-clinic-the-ai-survivor-penalty/ (Accessed July 9, 2026)