AI Hiring Pulse — Week of 2026-05-25

📊 Pattern of the week — an Israeli workspace SaaS just stood up an AI org

monday.com (NASDAQ: MNDY, HQ Tel-Aviv) opened 19 active AI roles in five days, spanning at least eight distinct product lines — and every single one of them sits in Tel-Aviv. It is the headline story of this pulse not because the absolute number is huge — Amazon opened 122 new roles in the same window — but because of the shape, and because of where it sits.

The seniority distribution is the second tell. Of 19 roles, eleven are Lead / Tech Lead / Group Lead / Engineering Team Lead — more managers and leads than individual contributors. You do not staff that ratio if you are adding capacity to one existing team. You staff that ratio if you are simultaneously launching multiple parallel teams that each need an owner from day one.

The stage shape confirms the read. Of the 19 roles, 17 are Agent — the application layer that wraps a foundation model into something a customer uses. One is Ship (product / GTM) and one is Serve (a DevOps Engineer for their BigBrain platform). They opened zero roles at Data, Pretrain, Post-train, or Eval — no data plumbing, no fine-tuning, no model training, no eval. monday.com is consuming foundation models via API and pouring all marginal AI headcount into agent and application work on top.

A methodology note: monday.com was added to our tracked universe on 2026-05-29; the dates above come from their Comeet ATS's posting timestamps and reflect their side of the posting, not ours. Future pulses will use our own first-observed timestamps for delta detection.

And a cohort-scale note. Across the eight Israeli-HQ companies we currently track — monday.com, Gong, Snyk, Wiz, Island, Cyera, Pinecone, Wix — combined active AI role count is 50. monday.com's one-week build-out (19 active roles) is larger than Gong (9), Snyk (8), and Wiz (6) combined. It is also larger than the active AI books at Cyera, Island, and Pinecone combined. Israel has driven a wave of NASDAQ-listed AI exits and ramps over the last two years — Wiz, Cyera, Snyk, Pinecone — but this week's signal is that the next leg of Israeli AI commercialization will be legacy Israeli SaaS rewiring themselves around agents, not just net-new AI-native cybersecurity and infra startups.

🪞 What's on the other side of net-negative

The week closed net -149 AI roles — the first net-negative weekly we've recorded. But it is not an industry-wide retreat. It is four companies:

Four companies account for -148 net; the other ~290 tracked companies are flat (net -1). That distribution is the actual story. The "AI hiring is slowing" narrative this week is, almost literally, the story of four discrete corporate decisions.

Two of those four — Deloitte and BCG — are consulting firms, and 63 of Deloitte's 71 closes sit at the Agent station (i.e., the application layer that wraps foundation models). That is not "AI is cooling at consulting." That is the agent-practice specifically being unwound. The same week monday.com is staffing six explicitly agent-titled roles plus a broader AI org in Tel-Aviv, the two biggest consultancies in our universe are closing 86 agent roles in the US. The throughline: agent work is moving from billable engagements at consultancies to product engineering inside customers. monday.com is one such customer.

The other two — Microsoft and Uber — are different shapes. Microsoft's -50 is broadly distributed (43 Agent, 13 Ship, 13 Serve, plus scattered upstream closes), reading more like a quarterly portfolio reset than a strategic exit. Uber posted zero new AI roles all week against 38 closes — pure shedding, no replacement. Last quarter Uber was the top consumer-sector AI hirer in our index; this week consumer is net -62 sector-wide.

🚨 4 inflections this week

1. The consumer-feed war is on pause. Uber 0/38, Whatnot 0/11, Opendoor 0/8, Roblox 0/4, Reddit 0/2 — none of the consumer companies that drove last quarter's Agent surge opened a single AI role this week. Consumer sector net: -62. This is not active retraction; it's an industry-wide breath-hold. The Agent share of new flow dropped from last pulse's 57.7% to 51.8% almost entirely because the consumer wedge of Agent went silent.

2. Pretrain stays at ~1% of new flow, now confirmed by an independent tag signal. Pretrain is below 1% of new roles this week (0.9% by stage) — and the pretraining tech tag is at the same 0.9% of new roles. Two independent classifiers, same answer. Pretrain as a hiring category is now confined to a handful of frontier labs. Eval Gate, which last pulse flagged as anomalously thin at 1.4%, bounced to 2.8% this week — driven by Amazon (4), Handshake (3), and Google (3). Provisionally back to baseline.

3. Robotics + defense + auto remain net +21 against the broader contraction. Anduril +3, Skydio +5, Wayve +6, Aurora Innovation, Applied Intuition all opened new roles. None individually large, but the cohort is the only sector — apart from enterprise — to net materially positive this week. Skydio alone opened 11 new AI roles (net +5 after closures), pushing its active book to 23.

4. JPMorgan stable at the top of bank AI hiring, but the bulge-bracket pattern from last pulse held. JPM opened 31 (vs last week's 29). Goldman / BofA / Citi / Wells / Morgan Stanley / AmEx combined: again 0. Two consecutive weeks of single-bank concentration is no longer a one-off — it is the steady state.

🔮 Prediction watch

Predictions to grade

• monday.com's Agent concentration is structural, not transient. Prediction: of the next 10 AI roles monday.com opens, ≥8 are at the Agent station. Cashable at 2026-07-06 (four pulses out). If wrong, monday.com is silently building model-layer capability we cannot see; if right, the headline "monday.com is now a foundation-layer customer" stands.
• The Deloitte agent-practice contraction continues. Prediction: Deloitte's Agent close-count exceeds Agent open-count in at least three of the next four pulses. This week's ratio: 63 closed vs 20 opened (3.15:1). Cashable 2026-07-06. If the ratio inverts back below 1.0, the unwind read is wrong and Deloitte was just rebalancing.
• Consumer Agent stays paused. Prediction: combined Uber + Whatnot + Roblox + Reddit + Opendoor new AI roles next pulse ≤ 10. Last pulse: 12. This pulse: 0. The two-pulse trajectory is the alarming read.

Predictions tracked

• Salesforce upstream hires < 5 over four pulses (made last pulse, cashable 2026-06-22). "Upstream" = Pretrain + Post-train + Serve combined (the three stages that imply Salesforce is building model-layer capability, not just consuming it). Salesforce opened 19 new AI roles this pulse: 15 Agent + 2 Ship + 1 Eval + 1 Post-train (the lone upstream hire). Running total: 1 of 5. Three pulses to go.

📰 Reading list

1. monday.com in our index — the full 19-role snapshot. Worth eyeballing the title list directly; the product-line sprawl reads differently in raw form than in a summary table.
2. Deloitte's Agent-station close pattern — 63 Agent closes in five days is the cleanest single-company signal of the "agent-practice retreat" thesis. Compare to monday.com on the same page.
3. Consumer sector trend — for grading the pause. The last-4-week chart should show this week as a visible inflection.
4. /insights/saas-vs-labs — the cohort comparison shipped last week. Pattern is the same one monday.com is exemplifying: application-layer SaaS hiring more AI engineers than the labs they integrate with.