Sysco has done something almost no board has: it tied executive equity to the cost-out behind its AI program. On September 9 the company said achievement of structural cost-out targets has been added to its long-term equity performance program, in the same paragraph as a target of at least $500 million in AI-powered savings by fiscal 2029. That is the right instinct. But beyond calling its first $100 million "net of investment" on an earnings call, Sysco hasn't said what the savings are net of, and until your compensation committee answers that question, copying the idea means paying executives for a number nobody has defined.
Three things are still open in Sysco's own filings: how much of the baseline is ordinary cost-cutting relabelled as AI, whether the cost of running the AI, not just building it, is subtracted, and which dollars belong to a separate synergy target from a $29.1 billion acquisition whose integration executives also hold performance awards for. Every company that ties pay to AI savings will have to answer the same three.
What Sysco Actually Committed To
Sysco committed to at least $500 million of AI and technology net cost out by fiscal 2029, and tied executive equity to hitting structural cost-out targets. The 8-K exhibit frames it as a target "to be realized by fiscal 2029," with $100 million of in-year savings already built into fiscal 2027 guidance. The Barclays conference deck filed the same day uses the more precise wording: "$500 million of AI and technology efficiency net cost out."
At the Barclays Global Consumer Staples conference, CEO Kevin Hourican said "there are 30 initiatives that ladder up" to the total and "four topics are going to drive 70% of the value." Those four are truck routing (Sysco drives 4.5 million miles a week in the U.S.), merchandising and fill rates, indirect spend, and customer-experience and back-office automation. Interim CFO Brandon Sewell said Sysco spends more than $1 billion a year on indirect expense and plans to use reverse-auction software to rebid the indirect spend that isn't competitively bid today.
The governance scaffolding is real, too:
- An AI transformation office. Hourican said it is run by "one of our top leaders," who reports to the CIO and meets weekly with him and the CFO.
- A renamed board committee. In August the board turned its Technology Committee into the Artificial Intelligence Transformation & Technology Committee, which now meets monthly.
- Money already paid. A September 4 8-K disclosed a $700,000 one-time cash award to the Chief Human Resources Officer for leading "the Company's efforts to transform its Artificial Intelligence ('AI') efficiencies."
For scale: Sysco reported fiscal 2026 adjusted operating expenses of $12.025 billion, so $500 million is roughly 4% of the cost base. That is big enough to move margins, and big enough that how it is measured matters.
Why Paying for Cost-Out Beats Paying for Adoption
Tying equity to realized cost-out is a better design than what most companies do, which is either pay for nothing or pay for rollout. Credit Sysco for choosing a P&L outcome over an activity count.
Almost nobody has gone this far. Mercer reviewed proxy filings from 145 S&P 500 early filers: 48% cite AI accomplishments, 14% have turned that into a short-term incentive metric, and only 2% put AI into long-term incentive plans. Strictly, Sysco would not make that list: its disclosed equity metric is "structural cost-out," not AI, and Mercer counted only metrics that name AI or a comparable technology. Sysco introduced it to align the company behind the AI program, which makes the metric's breadth part of the definition problem below.
The usual alternative is worse. Pearl Meyer warns that a metric based on AI deployment "may reward the rollout of tools, even if those tools are not meaningfully used". We have seen that failure in public disclosures. M&T published three Copilot counts and none of them was weekly actives, and DBS's 30% productivity figure was a goal, not a measured result. A cost-out metric inside a three-year equity plan avoids both traps. It measures money, and its horizon matches how long AI programs take to pay back.
So the design is sound. The definition is the weak point.
Chewy Took the Opposite Position the Same Day
On the same day, Chewy's CEO put about $50 million of annualized AI savings on the record and said it would not flow directly to the bottom line. On Chewy's second-quarter call, Sumit Singh said AI would save "low-tens of millions of dollars" in fiscal 2026, rising to "approximately $50 million on an annualized basis in fiscal 2027." Then he warned that "it would not be appropriate to mechanically layer" that on top of the margin trajectory. The savings will "help offset the normal cost pressures," including wage inflation, and some may be reinvested. His summary: AI is "a powerful enabler of continued margin not as a stand alone pool of savings."
Sysco's framing points the other way. On its fiscal fourth-quarter call, management said the team "has already voted to add structural cost-out to our long-term equity performance program" and that the improvement "will be leveraged to accelerate our debt reduction efforts and increase Sysco's overall profit margins."
Both positions are defensible, but they describe different things. Structural cost-out is a reduction in the recurring cost base that survives volume changes, wage inflation and reinvestment. It is not a one-time saving or a cost that was avoided. If Chewy is right that savings mostly offset wage inflation, operating expense doesn't fall. Under that definition, much of what companies call AI savings is cost avoidance. That is worth having, but it isn't structural, and it shouldn't pay out as if it were.
Net of What? Three Lines Sysco Hasn't Drawn
"Net" only means something once you say what gets subtracted, and Sysco's own disclosures leave three of those lines open. None of this suggests Sysco is doing anything wrong. It is an argument that the definition belongs in the grant, not the press release.
The baseline already includes cost-cutting that isn't AI
Sysco's first-year "AI" number includes a corporate cost program announced in April, before the AI savings target existed. On the Q4 call, management described fiscal 2027's figure as "approximately $100 million of in-year savings in fiscal 2027, inclusive of the cost-out savings we shared on our Q3 earnings call." On that April call, Sysco had identified $60 million of annualized run-rate savings from corporate expense optimization. The August release describes the $100 million as enabled by "AI-driven process improvements, automation initiatives, and operating efficiencies", and the deck says "AI and technology." Mixing them is fine for guidance. For a pay metric, you need to know which dollars the AI office actually produced.
The implementation bill may sit outside the adjusted numbers
In fiscal 2026, Sysco excluded $216 million of transformation costs from adjusted results, and it hasn't said where the AI program's build costs will be booked. The fiscal 2026 earnings release excluded "$216 million related to various transformation initiative costs, primarily consisting of supply chain transformation costs and changes to our business technology strategy," plus $71 million of restructuring, severance and legal costs. Routing software, procurement automation and back-office AI all fit that description. On the Q4 call, Sewell said the fiscal 2027 $100 million "is net of investment," but not what the investment includes or whether any of it sits in an excluded line. If it does, a savings figure measured on adjusted expenses has subtracted less than it appears.
Proxy advisors have a view on that. Pay Governance quotes ISS: adjustments "that appear to insulate executives from performance failures" count against a company. Run cost is the second half of the problem. Model calls, licences and vendor services recur every month, and Gartner warned in 2024 that CIOs could miscalculate AI costs by as much as 1,000% as they scale. A net figure that leaves out the growing per-task inference bill is really a gross figure.
A synergy target covers the same work
Sysco's pending Restaurant Depot acquisition has its own $250 million net cost synergy target, drawn from the same procurement and supply-chain work. The March announcement promised "approximately $250 million in annualized net cost synergies within the first three years following closing," coming "primarily via savings on product procurement and inbound supply chain optimization." The deal is expected to close by Sysco's fiscal third quarter of 2027. Two of the AI program's four areas are supply chain productivity and merchandising and procurement automation. Both targets appear in the same Barclays deck, each labelled "net," with overlapping three-year windows.
The pay structure doubles the problem. The September 4 8-K granted the CEO $2 million and the interim CFO $1 million in one-time PSUs that depend on the deal closing and pay out at 0% to 200% of target. The compensation committee said they are meant to keep leadership in place through the integration; the performance metrics were not disclosed. If a routing gain across the combined network can count toward both the deal PSUs and the structural cost-out metric, the same dollar gets paid twice. The fix is one attribution rule: every dollar of savings belongs to exactly one plan.
What Happened the Last Time a Savings Target Drove Pay Pressure
Once people are rewarded for a savings number, the bigger risk is no longer missing the target but inventing the savings to hit it. The clearest case is procurement. In 2021 the SEC charged Kraft Heinz, which paid $62 million. The SEC found its former COO "pressured the procurement division to deliver unrealistic savings targets," and the company restated $208 million of improperly recognized cost savings from nearly 300 transactions between late 2015 and 2018.
That isn't a claim about Sysco. The point is where savings are easiest to overstate: negotiated savings measured against a baseline the buyer sets. Savings from Sysco's planned reverse auctions on indirect spend are exactly that kind.
There is also turnover in the seats that measure the number. CIO Tom Peck left in April, and Navin Advani became interim CIO. That came two months after CFO Kenny Cheung announced he was leaving. The AI office reports to the CIO. The interim CFO presents the savings and holds one-time PSUs. None of that is improper, but it is why the definition should be written down independently before it is used for pay. Your company may have the same overlap; we saw a similar attribution problem when Group 1 cut 700 jobs and AI wasn't the reason.
Write the Definition Before You Write the Grant
If you plan to tie pay to AI savings, the metric definition is the deliverable. The grant comes second. Here is the sequence.
This Week:
- Tag every AI savings line in your last two board decks as gross or net, and as AI or non-AI. A line nobody can tag is not ready to go into a plan.
- Ask your controller which P&L line the AI build cost and run cost land in, and whether that line is excluded from the adjusted metric your incentive plan uses. That one answer tells you whether "net" means anything.
This Month:
- Write a one-page metric definition with five fields: the baseline period and how it adjusts for volume; what gets subtracted (inference and token spend, licences, vendor services, the AI team's own headcount); how reinvested savings are treated; how wage-inflation offsets are treated; and the attribution rule when an initiative overlaps a synergy or restructuring program.
- Have internal audit trace three claimed savings end to end, from invoice to contract to headcount record, before any number is used for pay. Start with procurement, where Kraft Heinz's savings were overstated.
- Report run cost per initiative monthly next to the savings it produces, so the net figure is calculated, not asserted. Hold it to the same standard as a holdout test for revenue claims.
Before the Next Grant Cycle:
- Have the compensation committee pre-approve the adjustment list at the start of the performance period, with a materiality threshold. Pay Governance suggests something like $5 million or 10% of target payout, so no exclusion gets decided after the results are in.
- Choose Sysco's position or Chewy's on purpose. If your savings mainly offset wage inflation, reward cost avoidance in the annual plan. Keep the long-term equity metric for cost that is actually gone.
The Bottom Line
Sysco's design is ahead of the market; its public definition is behind it. Every cost wave, from offshoring to shared services to cloud migration, has produced savings that were clear on a slide and hard to find on the P&L. The difference this time is that cost-out now sits inside executive equity, with a $29.1 billion acquisition's synergy target running alongside. Boards that copy Sysco should copy the structure, then write down the three lines Sysco hasn't drawn in public: the baseline, the run cost and the attribution rule.
A number you pay people to hit will get hit. Define it before it does.
Continue Reading
- DBS Put 1,500 Bankers on Agents. The 30% Is a Goal.
- M&T Published 3 Copilot Counts. None Was Weekly Actives.
- Airbnb Shipped 80% More Features. One Number Is Auditable.
- A Third Skipped a SaaS Buy. Now Price the Run Cost.
- Group 1 Cut 700 Jobs. Better AI Wasn't the Reason.
- Walmart and Amazon Both Said 40%. Neither Ran a Holdout.