Have you ever watched a market story flip so fast that last quarter’s doom narrative suddenly looks almost quaint? That is where enterprise software sits right now. For months, plenty of investors treated every customer relationship platform as if it were about to be eaten alive by chatbots. Then the mood shifted. Not because software stopped mattering. Because the companies that own the messy, permissioned, business-critical data started showing they can sit underneath the agents instead of competing with them for clicks.
The Panic Fades, The Model Gets Harder
I keep coming back to a simple observation. The so-called collapse of subscription software was never really about code disappearing. It was about fear that humans would stop opening apps, stop needing seats, and stop paying a predictable monthly fee. That fear had a catchy name in trading chats. It also had a weak grasp of how large companies actually work.
Large firms do not throw away a system of record because a model can draft an email. They worry about security, audit trails, who is allowed to see a quote, and what happens when an automated action touches a live customer. That is dull. It is also why the “everything goes to zero” story always felt overcooked to me.
What did change is the interface. People used to hunt through menus. Now they want to ask a question and let software do the walking. Salesforce spent a major customer gathering this week arguing that it can live in that world. The product pitch is clearer than the financial model. That gap is the whole story.
Why The Old Click Path Is Losing The Room
Think about a sales rep ten years ago. Open the account. Click related lists. Open a case. Open a campaign. Copy a number into a spreadsheet. Ask a manager what the discount policy actually is. Repeat that loop all afternoon. Training existed because the map of the product was the product.
The new loop is shorter and stranger. A person types a request into a conversation window. That window might live in a messaging tool, a general assistant, or the company’s own agent screen. The agent is supposed to know which records matter, which rules apply, and which systems sit next door. If that sounds tidy, it is not. It is a bet that the platform holding the data can expose logic without forcing every worker to become a power user.
Users no longer want a separate ritual for every application. They want one conversation that can reach several systems without a scavenger hunt.
In my view, that is the part Wall Street still underweights. The value is less “another dashboard” and more “the place an agent is allowed to act.” Once you frame it that way, the end of menu-driven work stops sounding like a funeral for software. It starts sounding like a redistribution of where software shows up.
Data Beats Screens When Agents Do The Work
Here is the unglamorous truth. A language model is only as useful as the context you feed it. Customer history, pricing rules, service entitlements, and approval chains are not sitting in a public crawl. They sit behind permissions. Salesforce’s argument is that those layers are the scarce asset.
That is why partnerships with general assistants matter more than splashy demos. If an outside model can reach Salesforce objects with the same security model a human would face, the company does not need to win every chat window. It needs to remain the trusted warehouse and the rulebook.
I’ve found that investors still talk about “apps” as if the logo on the login page is the product. For a lot of daily work, the logo is becoming costume jewelry. The product is the schema, the workflow engine, and the right to execute a change without creating a compliance mess.
- Customer records that already carry history and relationships
- Business logic that encodes how a company actually sells and supports
- Permissions that decide who, or what, may touch a field
- Connectors that let an agent hop across neighboring systems
None of that is poetic. All of it is hard to rebuild from a blank notebook. That is the quiet rebuttal to the idea that every firm will just spin up homemade tools and walk away from incumbents.
From One License Per Person To A Menu Of Deals
The seat was a beautiful invention if you were an investor who liked clean models. One person. One login. One recurring charge. Marginal cost near zero. Expansion arrived when a department hired more people or when a champion rolled the tool into another team.
Agents wreck that neatness. If software works while a human is in a meeting, who is the seat? If one digital coworker serves a whole service queue, do you price the bot, the tasks, the outcomes, or the humans who supervise it?
Leadership has been unusually blunt about customer taste. Some buyers still want per-user predictability. Some want to pay per agent. Some want raw consumption. Some want to pay when a transaction closes. A few even want a cut of savings or incremental revenue. That last bucket used to live in consulting slides. Now it is showing up in real conversations.
There is no majority bucket. Flexibility is the strategy, which is another way of saying the old template is no longer sacred.
I like the commercial honesty. I also see why analysts get twitchy. A sales force that can write “the best deal for that customer” is great for win rates and terrible for spreadsheet elegance. Mix shifts can hide inside a stable top-line number for a long time.
What Flexible Pricing Does To Margins And Models
Consumption sounds modern until you remember clouds already taught this lesson. Usage can surge. Usage can stall. Finance teams hate surprise bills, so they cap, commit, and negotiate corridors. Outcome pricing sounds even prettier until someone has to define the baseline. Saved money compared with what quarter? Incremental bookings after which campaign?
Perhaps the most interesting aspect is not which price wins. It is how many price books can coexist without turning the company into a custom shop. Software scale depends on repeatable packaging. If every contract becomes a science project, implementation cost creeps up even when product cost stays low.
| Pricing Style | Buyer Appeal | Modeling Pain |
| Per user | Familiar, budgetable | Low |
| Per agent | Matches new workflows | Medium |
| Consumption | Pay for activity | High |
| Transaction | Tied to completed work | High |
| Business outcome | Shares upside or savings | Very high |
None of this means growth dies. It means growth gets lumpier and harder to narrate in a single multiple. That is an overhang even when the strategic story improves.
The Stock Already Ran. The Debate Moved Upstairs.
Shares recovered hard from mid-year lows after results in late August forced a lot of short theses to shrink. Sentiment around the customer event was better still. The market is no longer pricing the firm as a relic. It is pricing a living franchise with a new set of questions attached.
That is a healthier argument. It is also a less generous one. When a stock is crushed on existential fear, a credible product tour can re-rate it quickly. When the fear is gone, investors start asking about mix, duration, and how much value the vendor keeps versus how much leaks to model providers and systems integrators.
Guidance did not get a theatrical lift at the investor session. Long-range revenue ambition stayed in place. That combination — better narrative, unchanged official numbers — usually produces a pause. People want proof in bookings quality, not another slogan about agents.
Customers Are Not A Monolith, And That Matters
Talk to enough buyers and you hear two honest, conflicting lines. One group likes seats because finance can forecast them and procurement already knows the dance. Another group expects agents to do more of the clicking and wants the invoice to follow work rather than headcount.
Both can be true in the same company. Marketing might keep named users. Service might buy capacity. A digital team might insist on usage. The vendor that can sell all three without a civil war in billing has an advantage. The vendor that pretends one metric will rule the decade is selling a sermon, not a contract.
In my experience, large software transitions look like this. Early adopters experiment. Core departments keep the old metric longer than Twitter threads predict. Then a renewal cycle arrives and the new metric shows up as an option, then a preference, then a default for net-new work.
Aggregation Is The Quiet Revolution
The line that stuck with me from the week was not about a feature name. It was about aggregation. For decades, companies tried to stitch sales, finance, and workforce systems together with projects that aged poorly. Agents offer a different glue. If several systems can be reached from one working environment, the agent becomes the integration layer people actually use.
That does not make integration cheap. It changes who feels the complexity. The user stops hopping. The platform has to interpret intent, respect entitlements, and write back safely. If Salesforce can be that hub — through its own coworker interface, through messaging, or through outside assistants — it stays in the path of work even when nobody “opens Salesforce” in the old sense.
Is that a revolution? Maybe. I’ve sat through enough enterprise keynotes to stay slightly allergic to the word. What I will say is this. If the interface becomes conversation, the companies that already store the nouns of the business are better placed than the companies that only store the adjectives.
Leadership Credibility Is Part Of The Asset
Markets love a founder who can sell a new chapter without sounding panicked. That matters here. The pushback against the extinction story was not subtle. It also happened to line up with how cautious enterprises buy technology. They experiment. They do not delete the ledger.
That does not make every product bet correct. It does mean the firm entered this cycle with a leader willing to talk pricing out loud instead of hiding behind “value-based transformation” fog. I’ll take blunt and incomplete over polished and empty.
Still, credibility is not a substitute for cohort data. The next few earnings calls will be judged on whether flexible packaging lifts win rates without chewing the contribution margin. If management can narrate mix with examples instead of slogans, the multiple can keep climbing. If every question about outcomes gets a shrug, the stock becomes a waiting room.
What “Thrive” Has To Mean In Practice
Thriving is not a press-release verb. For this company it probably means three unromantic things. First, remain the system where customer truth lives. Second, make that truth easy for agents to use without turning security into Swiss cheese. Third, get paid in a way customers accept and finance teams can still forecast.
- Keep data gravity inside the platform rather than leaking into one-off notebooks.
- Expose actions, not just answers, so agents can complete work.
- Offer more than one commercial motion without breaking operations.
- Show investors a repeatable pattern, even if the pattern is a mix rather than a single metric.
Miss the first two and the pricing debate is irrelevant. Miss the last two and the product can win while the stock sulks.
The Easy Money And The Harder Stretch
A rebound from deeply pessimistic multiples is the easy chapter. The market decided the franchise was not a melting ice cube. Fair. Getting from “not dead” to “deserves a growth software multiple again” is a different climb. That climb needs evidence that agents create more billable work than they retire in unused seats.
There is a version of this future where automation shrinks named users and consumption more than fills the hole. There is another version where every department buys digital coworkers the way they once bought extra licenses, and net revenue accelerates. Reality will land in the muddy middle. It usually does.
I do not think investors should treat unchanged long-term targets as a shrug. Stability after a narrative war can be a feature. It can also be a tell that management will not get ahead of a mix it cannot yet see. Both readings are allowed. The honest one is “we need more quarters.”
How To Watch The Next Year Without Getting Hypnotized
Ignore the mascot names. Watch four boring indicators. Attach rates of agent products on the installed base. Commentary on discounting when deals include new packaging. Gross margin comments that mention inference or usage costs. Customer language on whether they still budget seats as the default.
If attach is strong and seats are not collapsing, the hybrid model is working. If attach is loud and net expansion is quiet, you are watching a substitution story dressed as innovation. That distinction is worth more than any stage demo.
A simple field checklist: Are agents adding work or replacing licenses? Is pricing flexibility closing deals or confusing renewals? Does data remain gated, or is it leaking into unmanaged tools? Can finance still forecast the next four quarters?
Those questions sound pedestrian. They are the difference between a theme and a thesis.
A Personal Read On Terminal Value Versus Near-Term Friction
I am less worried about terminal value than I was a year ago. A mission-critical record system with permissions and process is not a toy. I am more worried about the awkward adolescence of pricing. Markets hate adolescence. They want a finished adult with a clean haircut and a single key performance indicator.
That impatience can cap the stock even when the product is right. It can also create entries for people who can live with a blurry model for a while. Neither side is silly. The mistake is pretending the blur is already gone.
Convincing the market that software still matters was the last job. Explaining how the invoice will look is the next one.
If that explanation arrives in pieces — a little more color on consumption here, a case study on outcome deals there — the multiple can grind higher. If it stays poetic, the shares may need actual beat-and-raise quarters to justify another leg.
What This Means For The Broader Software Complex
Salesforce is not a proxy for every subscription name. Horizontal tools with shallow data and weak workflow will feel more heat. Vertical tools that already sit in the transaction path may sneak through. The common thread is agent readiness. Can another system call you, act, and come back with an audit trail?
If the answer is no, you are a screen. Screens are getting cheaper. If the answer is yes, you might be infrastructure with a friendlier face. That is a better neighborhood, even if the rent formula is still being negotiated.
I’ve found that sector-wide multiples move on the loudest story, then sort themselves later. The loud story this year was extinction. The sorting story is monetization. Investors who skip the second chapter will keep confusing a product win with a financial one.
Risks That Do Not Fit On A Keynote Slide
Execution risk is obvious. Less obvious is cultural risk inside the sales organization. Reps trained on seat math need a new muscle. Deal desks need new guardrails so “flexibility” does not become a race to the most exotic discount. Partners need to know how they get paid when the hero of the story is an agent.
There is also model-provider gravity. If customers fall in love with a general assistant and treat the system of record as a quiet database, pricing power drifts. The counter is to make actions inside the record so valuable that the assistant is a doorway, not a replacement. Easy to say. Hard to productize across industries.
And yes, macro still exists. Software budgets are not immune to hiring freezes. An agent pitch that promises fewer people can freeze a deal as easily as it can close one, depending on who is in the room. Procurement hears “efficiency.” A line manager hears “headcount debate.” Same slide. Different pulse rate.
A Longer View Without The Cheerleading
Over a multi-year stretch, I still lean constructive. Not because every announcement will land. Because the firm is moving toward the way work is actually changing, and it is doing so from a position of data gravity rather than from a blank canvas. That is a better starting line than most critics admitted when the panic was fashionable.
The constructive case is not “easy money from here.” It is “the franchise survived a narrative that should have been more precise, and now it has to earn a new measurement system.” Surviving the narrative was necessary. Earning the measurement is the job that pays the next multiple.
So no, the death of software-as-a-service was not the plot twist of the decade. The plot twist is messier invoices, conversational interfaces, and a platform that wants to be used even when nobody is staring at its classic screens. That is less cinematic. It is also more like business.
If you came looking for a clean slogan, you will leave unsatisfied. Good. Clean slogans are how the last scare got too loud. The useful stance is quieter. Watch the mix. Watch whether agents create bills. Watch whether customers still treat this stack as the place truth lives. The rest is stage lighting.
And if the lighting is brighter than the contract language, wait. Markets eventually read the contract. They always do. The companies that get ahead of that reading, in plain speech, are the ones that turn a product adaptation into a durable valuation. That is the test now. Not survival. Translation.
