ACN › analyze
ACN · Analyze from before
A portfolio-specific passage was removed from the public build.
What this updates
The baseline was written one week before Q3 FY2026 earnings (June 18, 2026) and flagged that print as "the primary near-term risk." That print has now happened, and so has a second event the baseline could not have anticipated: an Aug-11 knowledge-base pattern note (pattern-headcount-revenue-divergence-tests-ai-deflation) that gives the central AI-bookings-vs-headcount question in the baseline's own guard a hard quantitative test, run across ACN's whole peer cluster (G, INFY, IBM, EPAM, CTSH).
Event list since the baseline (2026-06-11): 1. Q3 FY2026 earnings, June 18, 2026 — revenue $18.7B (+6% USD / +3% LC), bookings $19.3B (−2% YoY, −13% QoQ from the Q2 record), book-to-bill 1.0 (TTM 1.1), diluted EPS $3.80 (+9% YoY, beat), op margin 17.0% (+20bps). FY2026 guidance cut to 3–4% LC (from the "3–5%" the baseline carried forward from Q2). Stock fell sharply the next session (reported as a record one-day decline candidate). 2. Post-earnings capital-return escalation — buyback authorization raised $2B → total $7.5B FY2026 (+62% YoY) (June 23); acquisition investment plan raised to $9B for FY2026 (vs. $1.47B actually spent in FY2025). 3. $2.5B multi-year Microsoft agentic-AI pact and a ~€200M, 7-year NATO Protected Business Network contract (July 7) — both post-baseline. 4. Recent bolt-on acquisitions: McCoy (mid-market), Tokyo's COMWARE, a Dutch SAP transformation partner (Aug 25–27) — accelerating goodwill. 5. Sharp price recovery: low ~$118–147 in the June–July selloff window → $187–190 by late August, a ~28% rally off the July low. 6. Headcount: ~779,000 at FY2025 year-end (Aug 2025, after 22,000 "non-reskillable" role cuts, $865M restructuring cost) → ~799,000 currently (+2.6%), company guiding headcount to grow further in FY2026. 7. The Watchlist's own ACN line has already been updated once since the baseline report (conviction 7.0 → 6.5, entry revised to $130-145) — this pass tests whether that revision holds up under a full re-derivation, not just carries it forward.
The delta ledger
🔄 RETRACTED / needs correction
| # | Claim (baseline) | Status | What happened |
|---|---|---|---|
| 1 | "Q2 FY2026 (Feb 28, 2026): Revenue $18.7B vs est $18.53B; Adj. EPS $3.94 vs est $3.74; Adj. Op. Margin 17.0%" | ❌ RETRACTED | The primary source (Accenture's actual Mar-19-2026 Q2 FY2026 release) reports Revenue $18.04B (+8% USD/+4% LC) and EPS $2.93 (beat a $2.85 estimate, +4% YoY) — confirmed independently by Yahoo's own quarterly income statement for the Feb-28-2026 period, which reconciles exactly with these figures and with the TTM identity check (trailingEps 12.52 = sum of the last four quarterly diluted EPS values exactly, per the CLAUDE.md guard). The baseline's $18.7B/$3.94 figures do not match any Q2 print; they are closer to the real Q3 FY2026 print ($18.7B revenue, though EPS there was $3.80, not $3.94) — the baseline was written one week before Q3 reported and appears to have sourced or transcribed the wrong quarter. The one figure the baseline got right was bookings ($22.1B record — confirmed). This is a sourcing/transcription error in the prior report, not a vendor-feed trap — no correction is owed to a pitfall-* note, but it is owed to the record. |
🔄 SUPERSEDED
| # | Claim (baseline) | Status | What replaced it |
|---|---|---|---|
| 2 | "Record bookings $22.1B (Q2), AI bookings $5.0B+ — bookings trajectory... confirming demand is real and converting" | 🔄 SUPERSEDED | Q3 bookings fell to $19.3B — down 2% YoY and 13% QoQ from the Q2 record. Book-to-bill is 1.0 for the quarter (TTM 1.1, still >1). The Q2 record was a real number but was not sustained; one quarter of "record bookings" was not the trend the baseline's own tension-resolution ("watch closely if bookings growth slows — that's when the bear case materializes") anticipated it might not be. |
| 3 | "FY2026 guidance: revenue growth 3-5% local currency; ex-federal: 4-6%" (reaffirmed at Q2) | 🔄 SUPERSEDED | Guidance was cut to 3–4% LC at Q3 — the low end held, the top end came down a full point. This is the guidance-direction the baseline explicitly named as a break trigger (see row 15). |
| 4 | Fair value $195-225; Entry $155-175 (Watchlist since moved this to $130-145); Trim $250+ | 🔄 SUPERSEDED (Price rows never carry forward) | Re-derived below: FV $195-230, entry $155-175, trim re-expressed as 16x fwd (not a fixed dollar — CLAUDE.md rule). See Valuation. |
| 5 | "$130-145" entry zone (current Watchlist line, not the baseline report's own $155-175) | 🔄 SUPERSEDED | At spot $187.38 this zone is ~29-44% below spot and stale by the same mechanism flagged elsewhere in the Watchlist (pitfall-stale-entry-zone-suppresses-a-name). Restoring the baseline report's own $155-175 is the better-supported number once re-derived independently (below), not merely reverted. |
📉 DRIFTED
| # | Claim (baseline) | Status | Level at which it breaks |
|---|---|---|---|
| 6 | "The moat caveat: if Q3 bookings decelerate meaningfully (below $15B) or FY2026 guidance is cut to <3%, re-evaluate at conviction ≤5.5" | 📉 DRIFTED — tested, did not breach | Q3 bookings $19.3B (well above the $15B floor); guidance cut to 3-4% (floor held at 3%, not breached below it). Both of the baseline's own stated break conditions were tested by real events and both held — but both moved in the negative direction, which is the textbook DRIFTED case: moving against the thesis, not yet breaking it. Breaks at bookings <$15B or FY guide <3% in any future quarter — unchanged. |
| 7 | "98 of Fortune 100 clients... record bookings... AI bookings $5B+/quarter proves the bull case is currently winning" | 📉 DRIFTED | The bookings deceleration (row 2) and guidance cut (row 3) are the first data since the baseline that the "new AI demand is winning" claim has moved against, not with. Not broken — TTM book-to-bill is still >1 (1.1) — but the one-quarter trend that the baseline was leaning on to say "currently winning" no longer reads unambiguously that way. |
| 8 | "Zero sell ratings across 28 analysts for 4 consecutive months... mean price target ~$292" | 📉 DRIFTED | Current coverage: 25 analysts, average rating 2.07/5 ("Buy" but softer), mean target $184 — near spot, not 40%+ above it as the baseline's stale-target math implied. Two fresh upgrades post-Q3 (Wolfe $165→$215, Citi $135→$190, still Neutral) show analysts catching the rally, not leading it. The zero-sell fact likely still holds (not independently re-verified this pass — see Untested) but the target-based margin-of-safety argument the baseline leaned on has closed almost entirely. |
✅ CARRIED
| # | Claim | Check this pass |
|---|---|---|
| 9 | Capital-light model: FCF returned almost entirely to shareholders, minimal capex | Confirmed and strengthened. Buyback authorization raised to $7.5B (+62% YoY); YTD FY2026 FCF through Q3 is $8.78B (Q1 $1.51B + Q2 $3.67B + Q3 $3.60B), up ~24% over the comparable FY2025 9-month period ($7.06B) — FCF generation accelerated even as bookings decelerated. |
| 10 | Tangible ROIC ~60%, moat "real" | Not independently re-run this pass (no new annual statement since FY2025); nothing in the event list contradicts it. Carried on the strength of the prior quantitative work and no new counter-evidence. |
| 11 | Shares outstanding declining ~0.5-0.7%/yr via buybacks | Ordinary shares fell from 621.9M (Aug FY2025) to 611.9M (May FY2026 Q3) — -1.6% in three quarters, faster than the prior annual pace, consistent with the raised buyback authorization already executing. |
| 12 | Dividend grower: FCF payout ~34%, 3yr CAGR 15.2% | Current annualized dividend $6.52/sh (last quarterly $1.63×4) vs FY2025's ~$5.92 — +10.1% raise, a deceleration from the 15.2% 3yr CAGR but still a solid double-digit dividend grower. Payout ratio 51% of GAAP EPS (TTM) — richer than the baseline's FCF-basis 34%, but FCF-basis coverage remains ample given the FCF acceleration in row 9. |
| 13 | Structural moat sources (C-suite relationships, scale, backlog, global coordination) | No disclosed segment reclassification, no strategic redirection, no acquisition that changes what ACN sells. Structural claims hold by the standing prior — nothing in the event list rises to a business-model change. The McCoy/COMWARE/SAP-partner acquisitions are bolt-on scale, not a pivot. |
| 14 | Q2 FY2026 last quarter of separate AI-bookings disclosure | Confirmed. No standalone GenAI-bookings figure surfaces in any Q3 coverage found — consistent with the CEO's Q2 announcement that the metric would stop being broken out separately. |
| 15 | Goodwill risk: watch acquisition performance | Goodwill continued climbing: $22.5B (Aug 2025) → $25.3B (May 2026), and the FY2026 acquisition budget was raised to $9B — the trend the baseline flagged to watch is accelerating, not stabilizing. Still a non-cash / sentiment risk, not a cash-flow one. |
| 16 | DOGE/federal ~1% revenue drag, manageable | ⏳ Not independently re-tested this pass — see Untested. No fresh federal-specific figure was pulled for Q3; carried forward on the baseline's own reasoning only. |
🆕 NEW — no baseline counterpart
| # | Finding | Source | Direction |
|---|---|---|---|
| 17 | Headcount-revenue divergence test (the central AI question, quantified for the first time on ACN). Headcount ~779,000 (Aug 2025) → ~799,000 (current) = +2.6%. FY2026 guided revenue growth is 3-4% LC. That is a +0.4 to +1.4pt divergence on a local-currency basis — thin, essentially flat, nothing like Genpact's +9.7pt or IBM Consulting's +9.8pt-look-alike (even accounting for IBM's segment-proxy invalidity). ACN sits closer to the Infosys branch of pattern-headcount-revenue-divergence-tests-ai-deflation (client keeps most of the productivity gain) than the Genpact branch (provider keeps it and grows), though op margin did expand modestly (+20bps) rather than being fully given back. |
This pass, cross-referencing headcount web data against the Q3 guide | Mild negative force on the "AI is a clean tailwind" side of the baseline's central open question — not a break, but the first quantitative reading, and it does not support the bull case cleanly. |
| 18 | Net cash position fell from +$3.30B (Aug 2025) to +$1.78B (May 2026) — debt roughly flat (~$8.4B) while cash was drawn down to fund the $9B acquisition pace and the $7.5B buyback simultaneously. Debt/Assets ~12.2%, still low, but the balance sheet is being worked harder to fund two large capital-return/M&A programs at once. | Yahoo quarterly balance sheet | Neutral-to-mild-negative — still healthy, but a trend to watch if FCF ever softens while both programs stay at this size. |
| 19 | Insider activity across the full period since baseline: every disclosed transaction is either an RSU grant or a sale (Sweet, Walsh, Clifford, Hogan, Burgum, Unruch, Sharma, Macchi, Egawa) — sales recorded in Jan/Feb ($236-290), Apr ($174-178), Jul ($160-165), Aug ($175.98). Zero open-market purchases found. | Yahoo insider transactions | Weak negative signal, consistent with the same pattern seen at G (a cluster-wide observation, not ACN-specific) — normal for mega-cap equity comp cash-outs, not alarming on its own, but it does not corroborate the bull case either. |
| 20 | Microsoft $2.5B agentic-AI pact and NATO ~€200M contract | WebSearch, corroborated across multiple outlets | Positive — concrete evidence the AI-demand side of the thesis is landing real contracts, not just narrative. |
⏳ UNTESTED this pass
| # | Claim | Why untested |
|---|---|---|
| 21 | DOGE/federal revenue drag still ~1% | No Q3-specific federal figure was pulled; the baseline's own number is carried forward unverified. |
| 22 | "Zero sell ratings, 28 analysts" — current sell count | Current data confirms 25 analysts and a softer average rating (2.07) but the exact buy/hold/sell breakdown was not re-pulled this pass. |
| 23 | Q3 AI-related bookings quantification (any residual disclosure) | Not found in this pass's searches; presumed folded into the "no longer separately disclosed" claim (row 14) but not affirmatively confirmed absent. |
| 24 | Tangible ROIC and ROIC-incl-goodwill for FY2026 | No new annual statement exists yet (FY2026 closes Aug 31, 2026, four days after this pass) — cannot be recomputed until the Q4/FY2026 print. |
How the close calls were decided
Row 6 (the moat caveat / break triggers) was the central adjudication. Both of the baseline's own explicit downgrade conditions — bookings <$15B, guidance <3% — were tested against real Q3 data and neither breached. That is a genuine, mechanical test the baseline set for itself in advance, and it survived. Weighed against that: bookings and guidance both moved in the direction the bear case predicted (down), just not far enough to cross the line. The forces balance toward DRIFTED, not RETRACTED or SUPERSEDED — the thesis is not broken, but it is not gaining ground either, and the next print (Q4/FY2026, ~Oct 1) is the one that resolves whether 3-4% LC is a floor or the start of a slide toward the <3% break level.
Row 17 (headcount-revenue divergence) was weighed carefully because it is new evidence, not a re-test of an old claim. A single quantitative read is, per this command's own rule, "touched, not tested" — corroboration would require a second quarter's data or the company's own productivity-index disclosure, which ACN does not publish (unlike Genpact). This is recorded as a mild negative force, not a verdict — it does not by itself move ACN into the "loser" branch of the pattern note, because op margin still expanded (+20bps), which is inconsistent with a pure give-back story. It sits genuinely between the two branches, and that ambiguity is itself the finding.
Row 1 (the Q2 data error) was not escalated to a pitfall note. It is a sourcing/transcription error inside one report, not a repeatable vendor-feed defect — the correcting mechanism here is TTM-identity verification (sum four quarters, check it reconciles), which the CLAUDE.md guard already prescribes and which is what caught it this pass.
Thesis persistence and conviction delta
Structural + Trend claims tested: 10 (rows 9-16 plus the untouched ROIC claim). Carried/Refreshed: 8. Drifted: 2 (rows 6, 7 partially structural-adjacent). Retracted: 0 structural claims (the one retraction was a State-level earnings figure, not a structural or trend claim).
Thesis persistence ≈ 80-85% on Structural/Trend rows — high. The business model, capital-light returns, dividend growth, buyback pace, and ROIC characterization all held or strengthened. This is the classic "business held, multiple moved" pattern the framework watches for — except here the multiple moved up (11.3x → ~13.5x current-FY), not down, closing most of the valuation gap the baseline was pricing.
Conviction: 6.5 (was 7.0 in the baseline report). The Watchlist already carries 6.5 — this pass confirms that level rather than newly discovering it, via full independent re-derivation rather than a scan-level update. The 0.5-point move from the baseline's 7.0 is driven by three rows, in order of weight: (1) row 3/6 — guidance cut from 3-5% to 3-4% LC, testing but not breaching the baseline's own downgrade trigger; (2) the valuation cushion closing (price +11.8% against a fair value estimate that itself barely moved — see below — so the margin of safety that anchored the baseline's conviction is most of the way gone); (3) row 17 — the headcount-divergence test, the one hard quantitative read on the central AI question, comes back ambiguous-to-mildly-negative rather than supportive. Against this: the FCF acceleration (row 9), the raised buyback (row 9), and the Microsoft/NATO wins (row 20) are real positive forces, which is why this is a 0.5-point drift, not a full-point downgrade.
What is genuinely new (see rows 17-20 above)
The headcount-revenue divergence test (row 17) is the most consequential new finding — it is the first time the baseline's own central open question ("is GenAI a tailwind or a threat to billable hours") has been given a number instead of a narrative. The read is neutral-to-mildly-negative, which argues against treating ACN as automatically belonging to the "AI winner" side of its peer cluster the way Genpact does, and against automatically belonging to the "AI loser" side the way Infosys's four-year flat-margin story does. ACN's own data does not yet resolve its own question — which is itself the finding, and the reason conviction moved a half-point rather than staying flat or falling further.
Updated verdict
HOLD — conviction 6.5/10 (unchanged from the current Watchlist level; this pass independently re-derives and confirms it rather than merely carrying it forward).
Accenture remains a capital-light, high-ROIC dividend grower with a real (if structurally pressured) moat, and nothing in the event list since June breaks that. But the specific bull case the baseline built — "the market has priced in structural impairment that the numbers don't yet support, so this is deep value" — has substantially played out: the stock is up 11.8% while fair value moved almost nowhere (see below), so most of the identified mispricing has closed. What remains open is the AI question itself, and the first hard quantitative test of it (row 17) does not clearly resolve in ACN's favor.
Valuation (re-derived from scratch, current data)
| Model | Value | Weight | Weighted | Basis |
|---|---|---|---|---|
| Graham IV | $121 | 10% | $12.1 | √(22.5 × $12.52 TTM EPS × $52.12 BVPS) — floor only, punishes asset-light services |
| DDM | $176 (base; range $141-235) | 35% | $61.6 | D₁=$7.04 (8% growth off $6.52 current div, below the 15.2% 3yr CAGR to reflect deceleration), r=10%, g=5-7% |
| DYT | $238 (range $217-326) | 35% | $83.3 | Current yield 3.48-3.59% vs. 5yr-avg 1.64%, historical normal 1.5-2.5% |
| Bogle | $210 | 20% | $42.0 | Yield 3.5% + ~6% earnings growth + modest further re-rate toward 18x (already re-rated 11.3x→13.5x this year) |
| Weighted FV | ~$199 |
Fair value range: $195-230 (essentially unchanged from the baseline's $195-225 — the dividend grew, the multiple re-rated, and the two roughly offset). Current price $187.38 sits inside the low end of fair value, not clearly below it — the deep-discount setup the baseline traded on is largely gone.
A note on the forward multiple (per pitfall-vendor-forward-eps-is-the-wrong-fiscal-year, which fired on TSM/QCOM/BSX/AMD/ANET this month): Yahoo's forwardPE (12.79x) is built on epsForward $14.66, which is FY2027 (nextFiscalYearEnd = Aug 2027 vs. the fiscal year that closes in four days). The current-fiscal-year (FY2026) multiple is priceEpsCurrentYear = 13.52x off epsCurrentYear $13.86 — verified by summing three reported FY2026 quarters ($3.54+$2.93+$3.80=$10.27) plus an implied Q4 of ~$3.59, consistent with the guided range. Both numbers are legitimate; they are simply different years, and this report uses the FY2027 figure ($14.66) for the trim multiple below because "forward" in this repo's convention means next fiscal year.
Entry: $155-175 (restoring the baseline report's own zone via independent re-derivation, not reverting the Watchlist's stale $130-145, which sits 17-44% below current fair value and was never re-tested against the current data). Trim: 16x fwd (FY2027 basis, $14.66 EPS → implies ~$235, near the top of the FV range; converts the baseline's stale $250+ fixed dollar into the required multiple form, capped below the 20-25x historical-fair range to reflect the structural AI-era discount the Moat side of this file has carried since June). Hard stop: unchanged at $145 — a fundamental-deterioration level, not a sentiment one.
Break triggers (refreshed)
- Bookings fall below $15B in any quarter, or
- FY2026/FY2027 guidance is cut below 3% LC growth, or
- A second quarter's headcount-revenue divergence reads clearly negative (revenue growth falling meaningfully below headcount growth) — new trigger, added this pass, operationalizing row 17.
Upgrade conditions
- Two consecutive quarters of bookings re-acceleration back above the Q2 record pace, or
- The headcount-revenue divergence turns clearly positive (Genpact-shaped) for two consecutive quarters, or
- Price re-tests the $155-175 entry zone without a fundamental deterioration — a pure sentiment discount reopening the value case.
What this pass did NOT test
- DOGE/federal drag (row 21) — carried on the baseline's own unverified assertion; needs a Q4/FY2026 federal-specific figure.
- Analyst sell-count (row 22) — confirmed the average rating softened and the mean target compressed toward spot, but did not re-pull the exact buy/hold/sell tally.
- Residual AI-bookings disclosure (row 23) — inferred absent from silence in Q3 coverage, not affirmatively confirmed.
- FY2026 ROIC (row 24) — cannot be computed until the FY2026 annual statement closes (four days after this pass) and is filed.
- The headcount-divergence read (row 17) is single-quarter-equivalent evidence (year-over-year headcount change against one guided range) — per the pattern note's own rule, this needs a second independent quarter before it can move from "mild force" to a load-bearing claim. Flag this explicitly for the Q4/FY2026 print.
Sources: python .mcp/fin.py ACN --news, Yahoo Finance MCP (get_stock_info, quarterly income/balance-sheet/cash-flow, recommendations, insider transactions), WebSearch (Accenture Q2/Q3 FY2026 earnings releases via SEC 8-K/newsroom.accenture.com, Sherwood News, TIKR, ad-hoc-news.de, GuruFocus/NATO coverage, layoff trackers), Knowledge/Playbook/pattern-headcount-revenue-divergence-tests-ai-deflation.md, Knowledge/Playbook/pitfall-vendor-forward-eps-is-the-wrong-fiscal-year.md, Knowledge/Themes/consulting-it-services.md, baseline report Output/Stocks/Technology/ACN/analyze-2026-06-11.md, current Watchlist.md line 199.