| Last Week in HR AI: Issue #7 | Week of August 10, 2026 |
Cisco arms 90,000 employees; Rippling halves its AI bill
Accenture's own data shows reportable value slipping to 23%. The gap is foundations, not ambition.
|
1 TechCrunch2 ciodive.com3 forkast.news4 cities-today.com
Ikona's Take on the past week (August 10, 2026) in HR + AI
|
IO
|
Ian O'Keefe
Co-founder & CEO, Ikona Analytics
|
The economics of agentic AI stopped being theoretical this week. Rippling found its token spend had climbed to nearly 40% of the R&D compensation budget, driven by a small slice of employees, and responded by building a spend console, negotiating caps, and routing work to cheaper models. Volume held; cost fell to about 15%. Cisco is pushing a personal agent to all 90,000 employees with a CFO who talks about model routing the way he would talk about cloud commit. Both companies figured out how to meter the input. That is real progress, and it is the easier half of the problem.
The harder half shows up in Accenture's Pulse of Change data: agentic AI is beating productivity and satisfaction expectations, yet the share of firms reporting business value dropped to 23% in July from 32% earlier this year. Read those two facts together and the theme for the week is uncomfortable but clear: we are getting good at controlling what AI costs and still bad at proving what it returns, and the difference sits in foundations, captured judgment, workflow groundwork, and redesigned roles rather than in the models themselves. Raleigh is the proof in the other direction. Two agents now resolve close to half of internal IT tickets, but that number rests on more than a decade of data and workflow work that happened before any agent was switched on. The i4cp research points the same way inside HR: headcount is flat, the redesign is internal, and the firms pulling ahead are rebuilding HRBP, COE, and shared services around AI-enabled work instead of waiting for a headcount event to force the question.
So the question I would put to your team this week is not how much you are spending on AI. It is whether you can name the three workflows where an agent is carrying end-to-end work, and whether the judgment those workflows depend on exists anywhere outside the heads of the people who will retire (or worse, leave for a competitor) in the next few years. Wide and shallow produces local wins and no P&L line. If you are wrestling with where your own foundations are thin, reply and let us know. This is the problem we built Ikona to solve.
| The six to read | The stories that matter for the Office of HR — with our take on each. |
Rippling found that AI token spending had grown to roughly 40% of its R&D compensation budget, with a small fraction of engineers responsible for most of the consumption. The company built an internal AI Spend Console that pairs per-employee spend with output quality, negotiated caps with model providers, and added a gateway that routes work to cheaper models. Token volume stayed flat while total cost fell to about 15% of the R&D budget.
Cisco is deploying a personalized AI agent to all 90,000 employees, moving out of pilots and into operational integration. CFO Mark Patterson describes cost-disciplined model routing rather than defaulting to frontier models, alongside a "CFO cockpit" dashboard that already drafts most of the language in financial filings. The rollout runs alongside 4,000 job cuts scheduled for May 2026, AI-related orders rising from $2 billion toward $9 billion in guidance, and a stock up roughly 52% year to date.
Also covered by: fortune.com
Accenture's Pulse of Change survey of 3,000 C-suite leaders and 3,000 employees reports that agentic AI is exceeding expectations on both productivity and employee satisfaction. Even so, the share of firms able to report business value fell to 23% in July, down from 32% earlier in 2026. Muqsit Ashraf attributes the gap to wide, shallow rollouts that generate local gains without enterprise profit impact, and recommends taking a handful of use cases fully through to measurable outcomes while fixing data foundations, governance, and role design.
The piece introduces "tenure capital" to describe the pattern recognition and judgment senior employees build but never write down, framing it as an unmanaged asset ahead of a retirement wave in wholesale distribution. Drawing on Census Bureau aging data, Panopto research on knowledge loss, and remarks from a June symposium, it argues that AI agents placed at decision points such as rebate approvals and credit holds can capture the reasoning behind calls before people leave. It also disputes the idea that a younger, tech-fluent workforce solves the problem, contending AI multiplies existing judgment rather than rewarding tool comfort.
Raleigh, North Carolina is the first city government running ServiceNow's L1 AI Specialist in production, using two agents named Ral-E and Alli to autonomously resolve close to half of internal IT tickets, with a target of 85 percent. The deployment sits on more than a decade of data and workflow groundwork. The city plans to extend agents into resident-facing services, finance, facilities, and permitting, and its CIO frames the agents as trainees rather than replacements.
i4cp research covering 1,338 leaders finds that most large organizations held HR headcount flat over the past year, running counter to predictions of AI-driven reductions. The meaningful change is structural: future-ready firms are redesigning HRBP, COE, and shared-services roles around AI-enabled work, and shared services functions at top performers are twice as likely to be growing. Allianz Life is cited as an example, building AI upskilling into its workforce planning to reskill ahead of displacement.
Also covered by: aijourn.com, Human Resources Director
The argument is that treating AI delay as prudence is a mispricing: waiting compounds four costs, including tenured knowledge walking out the door, shadow architectures spreading unchecked, vendor commitments signed against an unknown current state, and foundation work that gets more expensive the longer it sits.
Raleigh's CIO framed it best: treat the agents as trainees, not replacements. Trainees need documented reasoning, clean workflows, and someone accountable for their output, which is exactly the work most organizations skipped on the way to the pilot. That work is unglamorous and it compounds, and the CHROs doing it now are the ones whose 2027 business cases will survive a finance review. Keep making the deliberate moves. This is the year they start paying off.
If your team is wrestling with workforce intelligence, AI readiness, or a transformation that has to land, we'd love to help. Reach out and tell us what you're working on.
Get in touch →Last Week in HR AI
Practitioner perspectives and diagnostic frameworks, delivered the week they publish. Be the first to know when new research drops.
No spam, ever. Unsubscribe anytime.