CBIZ Inc. ($1.7B mkt cap, 3 analysts) trades at 7x forward EPS and 14% FCF yield after a -57% YoY decline that consensus reads as "AI disrupting accounting." The Q1 2026 10-Q (filed 2026-04-30) and earnings call disclosed concrete AI/offshoring metrics that — per a 7-ticker accounting/advisory cohort cross-check — no peer is matching with comparable specificity. The thesis is misclassification; ≈40% of the idio alpha resolves on the July 29 Q2 print, the remaining ≈60% on multiple-rerate durability over 12-18mo.

What the filing says

Q1 2026 revenue $848.6M (+1.3% reported, Financial Services same-unit +1.8%). Management explicitly quantified Marcum integration headwind at ≈200bps on Financial Services organic growth — true underlying organic ≈4%, expected to abate in H2 2026 as integration-related attrition laps. Benefits & Insurance revenue declined -4.2%, attributed on the call to two isolated events (one producer team departure in February + tough contingent-commission comps); recurring B&I grew ≈4%.

Earnings call (2026-04-29): FY2026 FCF guidance $270-290M REAFFIRMED, EPS raised to $4.00-$4.10. Net leverage 3.4x, declining toward <2.5x by 2027. Management deployed $58M in repurchases through April 27 at $25-30/share, with the CEO calling shares "highly accretive" and "quite undervalued." Board renewed 5M share authorization Feb 2026 at multi-year lows.

The AI disclosure is what makes this memo. The 10-Q and call both quantified: offshore hours from 6% (2025) → 10% (2026) → >20% (long-term); 20% efficiency gain in tax workflows already achieved, targeting 40%; agentic AI company-wide rollout began Q2 2026. New CITO (Peter Scavuzzo) hired.

The most important bear data point: integration costs accelerated in Q1 ($23.8M vs $15.7M Q1 2025, +52% YoY). This is the wrong direction for the FCF recovery thesis. Management attributed it to busy-season front-loading + Marcum working capital settlement finalization, expects abatement H2. The bet is that Q1's $23.8M is busy-season front-load, not new run-rate. If Q2 stays at $20M+, FY2026 integration runs ≈$95M (vs $89M FY2025) — FCF guide breaks, multiple stays compressed, thesis damaged. Q2 print is the disambiguator.

The FAS reporting unit's $1.68B goodwill passed a Q1 quantitative impairment test (triggered by reporting unit reorganization) with a 7.4% fair-value cushion — ≈$124M buffer. A 15-20% sustained stock decline or 50bps WACC rise plausibly triggers a non-cash impairment of $200-500M, equivalent to $3.65-9.10 per share against a $1.9B equity base. Bears get sentiment damage even though the cash mechanics are unaffected.

What the market thinks

Forward P/E 6.88. Stock at 12% of 52-week range. RSI 51. Volume 0.4x 3-month average. Short interest 9.9% with 4.0 days to cover. Options structure unusual: ATM call IV 76% vs put IV 51.5% (calls 24.4% ABOVE puts — normal market shows put skew), P/C OI 0.27, IV rank 78%.

Risk-neutral probability of a $40+ print by mid-Oct (Oct 16 expiry, 159d, IV 60%): ≈28-32%. Our bull-state probability is 0.55 → an analogous P(>$40 within 12-15mo) is ≈50% — roughly the same construct viewed through different time windows, not two independent estimates. The disagreement is real, it's just one disagreement: market is pricing a smaller probability of multiple-rerate than we are.

Why the gap exists

Cross-ticker corroboration (Q1 2026 prints plus most recent calls) found CBZ alone in its accounting/advisory cohort with concrete AI/offshoring disclosure granularity. HRB silent. FCN silent. CRAI silent. ICFI silent. HURN: one vague "India managed services build" mention. KFY: abstract framing only ("agentic AI massive sucking sound") without offshoring % or named workflow efficiency metric. (KFY caveat: their last print was Sep 2025, FQ1 2026 — their next call mid-2026 is the real disambiguator on whether they retreat to abstraction or quantify like CBZ.) Adjacent IT/BPO peers (EXLS, ACN, Genpact, CTSH) are deep on agentic AI — but that IS their sector.

Three structural reasons the gap persists: (1) only 3 sell-side analysts cover CBZ; (2) CBZ's idiosyncratic variance is 51% against the peer-cohort regression — highest in the cohort, making it the alpha vehicle rather than a basket constituent; (3) forced-seller dynamics — Marcum shareholders receive stock monthly through ~early 2028 and have personal liquidity needs, supplying float absorbed by management's ROFR program plus open-market buybacks. The buyer is informed (management, plus CFO Form 4 purchases Aug+Nov 2025); the sellers are forced or narrative-driven.

Risks (ranked)

1. Q2 2026 integration costs stay >$20M (≈30%): FCF guide breaks. Largest near-term risk. Resolves at July 29 print.
2. FAS goodwill impairment (≈18%): $200-500M non-cash but sentiment-killing. Cohort has similar concentration (HURN 49.5%, ICFI 61%, CRAI $72M accumulated impairments) but no triggering events at peers. WACC sensitivity unmodeled — open research gap.
3. Project-based advisory softness: Management explicitly cannot forecast H2. Real but small impact relative to recurring base.
4. ASA fee revenue -6.4%: Mechanism unclear — could signal Marcum CPA partner/client leakage or internal revenue migration. Open research gap.
5. Cohort convergence on AI disclosure (≈30%): If 1+ peer matches CBZ's disclosure granularity, the discriminator advantage compresses.

Catalysts

• 2026-07-29: Q2 2026 earnings (binary on integration cost trajectory + AI disclosure sustenance — ≈40% of idio alpha resolves)
• 2026-08-05 to -15: Q2 10-Q filing (full disambiguation)
• 2026-12-31: Cohort peer-disclosure deadline (≈30% probability ≥1 peer converges)
• 2027-02-28: FY2026 10-K (FCF guide, buyback deployment, goodwill assessment)

What would change our mind

• Insider Form 4 OPEN-MARKET sells (not grants/tax)
• Q2 integration costs >$25M (vs Q1 $23.8M) — confirms acceleration not abatement
• Management retreats from concrete AI/offshoring metrics to abstract framing on Q2 call
• 2+ cohort peers (HRB/HURN/CRAI/ICFI/FCN/KFY) disclose concrete offshoring % in next earnings cycle
• Goodwill impairment 8-K preceding 10-Q
• FCF guidance withdrawal or cut

Forward EV — and the honest version of the bet

Probability-weighted 12-15mo scenario tree: bull $50 (P=0.55), base $36 (P=0.25), bear $26 (P=0.15), tail $18 (P=0.05). E[r] +28.7% over 12-15mo, ~+23% annualized. Idio Sharpe 0.48.

Be clear on what this is. Strip the bull state and E[r] ≈ 0%. The base/bear/tail legs collectively contribute roughly nothing. The trade is a probability-weighted bet on the bull state materializing within 12-15mo. The bull state requires both Q2 confirming integration cost decline AND the cohort-discriminator narrative actually moving the multiple over 12-18mo. Either failing collapses the EV to flat or negative.

Position structure: starter size within management's buyback band; scale only on Q2 confirm. Pre-catalyst sizing intentionally small because the dominant driver requires the binary disambiguator to materialize.

Evidence

Independence note: rows 1-5 below all source from the same Q1 2026 earnings call (2026-04-29) and should not be multiplied as independent signals — combined LR ≈ 1.5 (max), not the product (≈6.6×). Rows 6-12 are independent.