Salesforce, Inc.: Can Agentforce Reaccelerate CRM Growth?

I usually spend more time on small caps. I look for the point where technological change and financial statements meet; if I need a label, “techno-fundamentalist” is close enough. So when large SaaS stocks moved into lower valuations, my first question was not whether they looked cheap. It was whether AI had made them easier to reproduce.

Cheaper software creation can erode a moat. If an AI agent does the work of several users, a company that charges by seat may improve the customer’s productivity while reducing its own license count. That concern lands directly on Salesforce, Inc. (NYSE: CRM).

But Salesforce is not a startup rebuilding CRM on a blank page. Customer relationships, permissions, sales and service workflows, and enterprise data already live inside its products. If Data 360 unifies the data and Agentforce acts on it, the installed base becomes distribution. If one agent replaces several seats, the same installed base becomes the pool being cannibalized.

I am less interested in whether Salesforce can earn the label “AI company” than in how much AI success today’s price already requires. That reverse-pricing question follows the logic of my Sterling Infrastructure valuation; the question of who captures the economics from an AI feature connects to Rogo and financial AI.

CRM
customers→Data
360→Agentforce→AI
usage

+ usage, upsell, new work

− seats and license cannibalization

1. Salesforce is becoming an AI company

Sales Cloud holds the pipeline, Service Cloud runs support workflows, Slack is a surface on which people and agents work, Data 360 connects records, and Agentforce acts on that data under enterprise permissions.

The chain can run two ways. In the upsell path, an existing CRM customer buys Agentforce usage and Data 360, lifting revenue per account. In the substitution path, the agent handles prospect research, ticket triage, or response writing and allows the customer to buy fewer seats.

Salesforce’s defense is less the model than the data, permissions, deployment, and switching cost around it. A general model can be replaced. The rules deciding which customer gets which price, who may approve a refund, and what data an agent may touch still live in an enterprise system. That is a moat. It is not a permanent one: Microsoft, ServiceNow, HubSpot, and new AI applications are aiming at the same layer of work.

2. Eleven percent growth is not one kind of growth

FY2027 Q2 revenue was $11.345 billion, up 11%. Subscription and support revenue rose 12% to $10.820 billion. On the surface, growth is back in double digits.

Informatica contributed $456 million to total revenue and $440 million to subscription and support. Salesforce guided FY27 revenue to $46.1–46.4 billion, up 11–12%, and said Informatica adds slightly more than three percentage points. The guidance raise also includes the pending Contentful and Fin acquisitions. FY27 Q2 results

The useful approximation is:

$$g_{reported}\approx g_{organic}+g_{M\&A}+g_{FX}$$

The actual calculation is harder. It has to align acquisition dates, the prior-year comparison, currencies, and other transactions. The dotted series below is therefore not Salesforce’s official organic-growth metric. It is an author estimate that simply removes disclosed Informatica revenue from the numerator.

RevenuecRPOEx-Inf. est.

FY27 Q2: Revenue 11% · cRPO 14% · simple ex-Informatica estimate 6.4%

That estimate is about 8% in FY26 Q4, 8.8% in FY27 Q1, and 6.4% in FY27 Q2. It does not prove that the core business is decelerating; currency, deal timing, and other acquisitions remain. The narrower conclusion is stronger: the reported 11% should not all be read as a reacceleration of the underlying Salesforce business.

3. Does cRPO lead a recovery?

FY27 Q2 current remaining performance obligation was $33.5 billion, up 14%; total RPO was $66.3 billion. cRPO is the contracted amount expected to become revenue over the next 12 months. It is neither cash collected nor all future revenue.

Across the latest ten quarters, cRPO growth generally matched or ran ahead of revenue growth. In FY26 Q1, cRPO rose 12% while revenue grew 8%. In FY26 Q4, cRPO grew 16% including Informatica. It was still up 14% in FY27 Q2.

RevenuecRPOEx-Inf. est.

FY27 Q2: Revenue 11% · cRPO 14% · simple ex-Informatica estimate 6.4%

The lead is not mechanically one quarter or four. Renewal timing, contract duration, currency, large deals, and acquisitions all move the metric. Recent cRPO and reported revenue both include Informatica, so neither should be compared directly with an ex-acquisition series.

My conclusion is conditional. Sustained cRPO growth above revenue can be an early sign of reacceleration. It is closer to a necessary condition than a sufficient one. The test is whether it turns into ex-acquisition revenue over the next four quarters.

4. What does Agentforce ARR actually measure?

Salesforce reported more than $1.5 billion of Agentforce ARR in FY27 Q2, up more than 240%. Agentforce plus Data 360 ARR was about $3.9 billion, up more than 210%. Agentforce and Slack had processed 7.0 billion cumulative Agentic Work Units, including 3.2 billion in Q2, up 97% sequentially.

There is a consequential footnote. Beginning in FY27 Q2, Agentforce ARR includes additional AI offerings such as Slackbot and Headless 360. The 240% figure is not a perfectly like-for-like ruler across every period. ARR annualizes active subscription agreements; it is not recognized revenue, net-new revenue, or cash flow. Metric definitions

Economically, I separate four pieces:

1. AI usage purchased by existing customers
2. new customers and workflows won because of AI
3. CRM seats or licenses replaced by AI
4. incremental inference, serving, implementation, and support costs

$$\Delta FCF_{AI} =\Delta Revenue_{AI}\times ContributionMargin_{AI} -LostContribution_{CRM} -IncrementalFixedCosts$$

This is my framework, not a Salesforce reporting metric. The illustration below is also not guidance. With $5 billion of year-five incremental AI revenue, 15% cannibalization, and a 55% contribution margin after that lost revenue, net FCF contribution is $2.34 billion. Five billion dollars of ARR does not become five billion dollars of value.

$5.00BGross AI revenue

−$0.75BCRM cannibalization 15%

−$1.91BAI cost, tax, etc.

$2.34BNet FCF contribution

There are three plausible paths: AI Upsell preserves seats and adds usage; AI Substitution adds usage but reduces seats; AI Expansion wins work outside the installed CRM base. Current disclosures do not reveal the mix. That is why I do not turn ARR growth directly into a bullish valuation.

5. Free cash flow per share is the KPI that survives the story

Salesforce generated $15.0 billion of operating cash flow and $14.4 billion of FCF in FY26. Management guided FY27 FCF growth to roughly 4–5%. Applying the 4.5% midpoint produces an FY27 estimate of $15.05 billion. That figure is an author estimate.

At July 31, 2026, Salesforce held $8.310 billion in cash and $3.093 billion in marketable securities against $39.288 billion of long-term debt. First-half CFO was $7.970 billion and capex was $316 million. It spent $27.332 billion on repurchases and $729 million on dividends while raising $24.842 billion of debt. FY27 Q2 10-Q

Share count also has two definitions. Q2 diluted weighted-average shares were 821 million; actual shares outstanding were about 823 million on August 20. The $25 billion ASR initially delivered 103 million shares at $198.34, with final settlement still due in Q3. I use 823 million for a current per-share value.

Estimated FY27 FCF per share is about $18.28. At $224.56, the stock starts at roughly 12.3 times FCF. That does not look demanding. The question is what quality of cash flow sits in the denominator.

6. FCF can be overstated when SBC is ignored

First-half FY27 stock-based compensation was $1.763 billion, or $3.526 billion annualized. SBC is added back as a noncash charge in operating cash flow. It is not free to shareholders: new shares dilute them, or the company spends cash on repurchases to neutralize the issuance.

There are two coherent methods:

• Reported FCF plus share-count modeling: use reported FCF and model the net share count.
• SBC-adjusted owner earnings: deduct some or all SBC as an economic cost.

Using both in full double counts the same effect. My base DCF is closer to the first method: reported FCF and today’s 823 million shares. A conservative cross-check deducts annualized SBC. In the simulator, share-count change only affects the year-five FCF-per-share bridge. It does not also increase DCF equity value, because buybacks consume cash.

Reported FCF$15.05BFY27E

−

Annualized SBC$3.53B100% economic-cost lens

=

Conservative owner earnings$11.52Btax effects simplified

GAAP EPS also needs a bridge. Salesforce booked a $2.613 billion strategic-investment gain in FY27 Q2 and said it added $2.43 to diluted EPS. Q2 GAAP EPS of $4.29 was not all operating improvement. I do not add the $11.324 billion strategic-investment portfolio separately to the DCF; it is volatile, partly illiquid, and can create double-counting.

7. Valuing Salesforce

The primary model treats CFO − capex as an FCFE proxy. Because this cash flow is after interest expense, I discount it at the cost of equity and do not subtract debt again. It is not a fully specified FCFE model of new borrowing and principal repayment; stable capital structure is a limitation.

The model starts at FY27 estimated FCF of $15.048 billion, forecasts five years, and adds a Gordon terminal value. Share count is 823 million. Every scenario is an author estimate.

Scenario	5-year FCF CAGR	Discount rate	Terminal growth	Year-five FCF	Fair value
Bear	2%	11%	2.0%	$16.6B	$207
Base	7%	10%	2.5%	$21.1B	$302
Bull	12%	9%	3.0%	$26.5B	$459

Bear$207

Base$302

Bull$459

Market$224.56

The $302 base value is about 34% above the market price. The price has roughly a 26% discount to that value. A 20% margin-of-safety price is $241; 25% is $226; 30% is $211. The current price sits near the 25% line.

But 72% of base value comes from the terminal value. Small changes in discount and terminal growth rates move the answer. DCF is a scale for comparing expectations, not a price target accurate to the dollar.

FCF CAGR discount	8%	9%	10%	11%	12%
2%	$333	$282	$245	$216	$193
4%	$364	$308	$266	$235	$210
6%	$397	$335	$290	$255	$228
8%	$432	$364	$315	$277	$247
10%	$470	$396	$341	$300	$267

8. What does the market price require?

The most important calculation in this note runs backward:

$$P_{market}=DCF(g_{implied})$$

Holding a 10% discount rate, 2.5% terminal growth, and 823 million shares, the five-year FCF CAGR needed to explain $224.56 is about 0.01% — effectively zero. FY27 estimated FCF of $15.05 billion only needs to persist.

The sentence changes with SBC. Deduct 25% of annualized SBC as an economic cost and implied growth becomes 1.4%. At 50%, it is 2.9%. Deduct all of it and the required rate is 6.3%.

SBC 0%0.01%

SBC 25%1.4%

SBC 50%2.9%

SBC 100%6.3%

That is why Salesforce can look cheap and contentious at once. Reported FCF says the market requires almost no growth. A full owner-earnings treatment asks for growth somewhat above the company’s FY27 FCF guide of 4–5%.

I do not pick one extreme without evidence. If net shares keep falling, leverage stabilizes, and Agentforce adds cash flow, reported FCF is the more useful lens. If buybacks mostly absorb SBC while debt and interest remain, owner earnings is closer to the economic truth.

Change the assumptions directly in the Salesforce Valuation Lab. The reproducible model lives in src/data/salesforce-valuation.ts, the verification script in scripts/salesforce-valuation.mjs, and unit tests in tests/salesforce-valuation.test.mjs.

9. Conditions under which I turn more positive

At the current numbers, Salesforce belongs in a “consider buying” range. It is below my base DCF, and reported FCF embeds nearly no growth. It is also only about 8% above the $207 bear value. This is not an unconditional bargain; it is a price that works if cash-flow quality holds.

I would become more confident if:

• growth excluding disclosed acquisition effects returns to 8–10%;
• double-digit cRPO growth converts into revenue over the next four quarters;
• like-for-like Agentforce ARR and usage rise together;
• FCF margin holds as AI usage scales;
• leverage stabilizes and actual shares outstanding keep declining after the ASR;
• FCF per share rises after an SBC adjustment.

If ex-acquisition growth remains below 6%, FCF stays flat, and the share count stops falling into FY28, the 7% base assumption needs to come down. A price below $207 would not be “cheap” if the business thesis had broken.

10. How I could be wrong

I could be too conservative. Salesforce’s customer relationships and enterprise data may be a formidable distribution channel for Agentforce. Usage revenue may more than offset fewer seats. Slackbot and headless agents may capture work outside CRM. If FCF grows 12% and shares also decline, $224 will have been a very low price.

I could also be too optimistic. General models and new applications may take the interface and leave Salesforce as a lower-value system of record. Agentforce revenue may replace seats while inference, implementation, and support consume the economics. Definition changes make ARR continuity harder to verify.

The capital structure matters too. The $25 billion ASR reduced shares quickly but was funded in meaningful part with debt. Persistent leverage and higher interest can pressure FCF. If SBC remains high, part of buyback spending is dilution defense rather than value creation. Informatica and further acquisitions lift reported growth while adding integration risk and potentially obscuring organic weakness.

Conclusion: watch cash flow per share, not AI ARR alone

It is too early to declare that Salesforce’s organic growth has recovered. cRPO is a constructive leading signal, but acquisitions and contract timing affect it. Agentforce ARR is growing quickly, but its definition changed and Salesforce does not yet split net-new revenue, cannibalization, and inference cost.

The price is still interesting. Almost no reported-FCF growth is needed to explain $224.56, while a 7% base case produces $302. Under a full SBC adjustment, however, the required growth rises to 6.3%. The investment question is less “Will AI work?” than “Can AI and capital allocation produce 6–7% growth in FCF per share?”

I see the stock between undervalued and fairly valued — a level worth considering, not a small-cap-style asymmetry. It is the intersection of a large installed base, lower expectations, and strong cash generation. The proof will not be a bigger Agentforce announcement. It will be ex-acquisition growth and per-share cash flow after SBC.

A good company and a good stock differ by the price of expectations. Before deciding whether Salesforce becomes an AI company, calculate how much success the current price already demands.

Disclosure: This article records personal investment research and judgment; it is not investment advice. Fair values are author estimates based on assumptions, and actual outcomes may differ materially. The author’s CRM position status is not reflected in this draft.

Sources and reproducibility

The as-of date is October 8, 2026. Financial data prioritize Salesforce’s FY2027 Q2 10-Q and earnings release plus FY2026 full-year results. The share price is the October 7 regular-session close of $224.56, not the $224.44 after-hours quote. Quarterly time series come from Salesforce releases. Dotted chart lines, the AI waterfall, DCF, reverse DCF, and SBC adjustments are author estimates.