The Thesis
The enterprise AI cost structure is being set right now, through flat-fee agentic licenses, headcount-funded pivots, and federal regulatory deadlines that are repricing compliance, and operators who wait will pay more at every layer.
The Signal
Three moves worth watching this week
Adecco locks in unlimited agentic AI for a flat fee.
On March 12, the Adecco Group signed a multi-year Agentic Enterprise License Agreement with Salesforce for unlimited access to Agentforce 360, targeting more than 50% of revenue powered by agentic AI by year-end (Salesforce newsroom, March 12, 2026).
This matters because it is the first Fortune 500 deployment to treat agent capacity as a fixed operating cost rather than a variable experiment, converting AI from a line-item pilot into a capital-allocation decision.
The second-order effect: Gartner has warned that these agreements may convert to defined-quantity contracts at renewal, meaning the flat fee is an acquisition cost, and switching costs compound from day one (The Register, Jan. 27, 2026).Atlassian cuts 1,600 jobs to self-fund its AI pivot.
On March 11, Atlassian announced it would eliminate 10% of its workforce, roughly 1,600 roles, to redirect capital toward AI development and enterprise sales, taking $225M-$236M in restructuring charges (CNBC, March 11, 2026).
This matters because it follows Block (4,000 cuts) and WiseTech (2,000 cuts) in establishing "restructure to fund AI" as a standard board-level pattern; over 45,000 tech workers have been laid off globally in 2026 so far (Computerworld, March 12, 2026).
The second-order effect: Atlassian's stock is down 50%+ this year on fears that tools like Claude Cowork threaten its core collaboration products, meaning the restructuring is defensive as much as strategic, and CTO Rajeev Rajan's concurrent departure signals a deeper architectural reset (GeekWire, March 11, 2026).Federal AI regulatory deadlines arrived this week.
The March 11 deadline required both a Commerce Department evaluation identifying state AI laws the federal government considers "onerous" and an FTC policy statement on how Section 5 applies to AI models, with an operational DOJ AI Litigation Task Force ready to challenge state laws in court (Baker Botts / Mondaq, March 10, 2026).
This matters because operators now face a two-track compliance reality: state laws like California's and Texas's remain fully enforceable absent court action, even as the federal government signals it will sue to preempt them (Ropes & Gray, March 11, 2026).
The second-order effect: Colorado's AI Act (effective June 30) is the next flashpoint. Any company deploying high-risk AI systems needs to maintain parallel compliance tracks until a court rules, which could take months or years.
The Playbook
Agentic License Evaluation Rubric: 5 questions before you sign (or build)
Use this in your next vendor review or board-level AI budget discussion.
What is the all-in cost per automated workflow at Year 1 vs. Year 3? Flat-fee AELAs look cheap at signing. Model the renewal scenario. Gartner warns vendors may convert unlimited agreements to defined-quantity contracts. Calculate your per-workflow cost at current usage and at 3x usage to see where the economics invert.
What data does the vendor retain, and what are the connector fees? Constellation Research flags that enterprise data tolls and API economics are emerging as the largest hidden cost of scaling AI agents. Ask specifically: does the vendor charge for cross-cloud data movement? Do connector fees increase with agent volume?
Can you measure agent-attributed value today, not projected value? If you cannot tie a specific agent action (case resolved, cycle time reduced, lead qualified) to a dollar outcome, you cannot negotiate the renewal. Build the measurement system before you sign the license, not after.
What is your switching cost at 18 months? Map every integration point. Count the data dependencies. If the AELA includes Data Cloud or MuleSoft, your exit cost rises with every connected system. Quantify this before the CFO sees the headline number.
Does your regulatory exposure change under this agreement? If you deploy AI agents in Colorado, California, or the EU, the compliance obligations may differ depending on whether the agent is classified as a "deployer" tool or a "high-risk system." Your General Counsel should review the agreement alongside your compliance map, not after it.
The Verification Test
Claim: "Our agentic AI platform delivers 3-4x ROI within the first contract year."
Test: Request a customer reference that has completed a full fiscal year on an AELA or equivalent flat-fee agentic license and can provide audited before/after metrics on at least one automated workflow.
Pass criteria: The reference can name the specific workflow, the baseline metric (e.g., average case resolution time, cost per qualified lead), the post-deployment metric, and the measurement methodology. The CFO or equivalent signed off on the numbers.
Fail smell: The vendor offers only "projected" ROI, references pilot results (sub-6 months), or provides case studies from internal deployments rather than paying customers. If the best available evidence is a conference-stage testimonial, the ROI claim is a forecast, not a finding.
HSI Note
Horizon Search Institute
Human Performance: U.S. House subcommittee held hearing on AI-ready workforce training under WIOA, signaling bipartisan pressure to retool federal reskilling. Link
Responsible AI: Ropes & Gray analysis finds federal preemption of state AI laws faces significant legal obstacles, meaning compliance fragmentation persists. Link
Planetary Futures: Lilly's LillyPod, the pharma industry's most powerful AI factory (1,016 Blackwell Ultra GPUs), is now operational for drug discovery at scale. Link
Governance and Diplomacy: EU Commission proposes delaying high-risk AI Act enforcement to late 2027, widening the transatlantic regulatory gap for multinational deployers. Link
Links Worth Your Time
Baker Botts, "March 2026: Federal Deadlines That Will Reshape the AI Regulatory Landscape" — The clearest legal briefing on what the Commerce and FTC March 11 deliverables mean for operators and where litigation risk sits next. Link
Forrester, "AI Agents Become Economic Actors" — Analyzes how Salesforce, Microsoft, and Google are repricing AI from usage volume to economic outcomes, and what this means for CFOs governing AI budgets. Link
CNBC, "Atlassian slashes 10% of workforce to self-fund AI" — Best factual coverage of the restructuring, including the $225M-$236M charge estimate and Cannon-Brookes's internal memo framing. Link
Constellation Research, "Enterprise Technology 2026: 15 Trends to Watch" — Larry Dignan's breakdown of agentic enterprise licensing, data toll economics, and why decision velocity may matter more than model capability. Link
Fortune, "Nvidia GTC Preview" — Pre-event briefing on what GTC 2026 announcements (NemoClaw, new inference chips, Groq integration) could mean for the enterprise AI infrastructure stack. Link
The Builders & Doers Podcast
Mike Wu’s résumé looks like the classic prestige path: USC, investment banking, Sony, Harvard Business School, BCG, and then entrepreneurship. But in this conversation, he explains why that neat story looked much cleaner on LinkedIn than it felt in real life.
Sources
Salesforce Newsroom, "The Adecco Group to scale agentic AI at speed with unlimited Agentforce license agreement," March 12, 2026.
The Register, "Gartner questions if Salesforce AI will stay all-you-can-eat," January 27, 2026.
CNBC, "Atlassian slashes 10% of workforce to self-fund investments in AI," March 11, 2026.
Computerworld, "Atlassian cuts 1,600 jobs to fund AI and enterprise expansion," March 12, 2026.
GeekWire, "Atlassian layoffs impact 63 workers in Washington as CTO steps down," March 11, 2026.
Baker Botts via Mondaq, "March 2026: Federal Deadlines That Will Reshape the AI Regulatory Landscape," March 10, 2026.
Ropes & Gray, "Examining the Landscape and Limitations of the Federal Push to Override State AI Regulation," March 11, 2026.
Forrester, "AI Agents Become Economic Actors: Salesforce Rewrites the Rules of Pricing," December 18, 2025.
Constellation Research, "Enterprise Technology 2026: 15 AI, SaaS, Data, Business Trends to Watch," 2026.
Fortune, "Nvidia GTC Preview: The 5th Layer of AI," March 12, 2026.
NVIDIA Blog, "Lilly AI Factory Live," February 27, 2026.
U.S. House Committee on Education & the Workforce, "Building an AI-Ready America: Strengthening Employer-Led Training," March 2026.
European Commission, "AI Act: Regulatory Framework," 2026.
The White House, Executive Order: "Ensuring a National Policy Framework for Artificial Intelligence," December 11, 2025.
The Searchlight is a weekly field memo. It is not investment advice. Views are the editor's own.