resources

The world is changing. Is your organization part of it?

In one year the platforms your company runs on rebuilt their pricing, their products and their partnerships around agents. A note to executives on the one decision that is still theirs.

September 13, 2026

10 min read

THE SHIFT

WHAT IT ASKS OF YOU

THE DECISION

The world is changing. Is your organization part of it? - cover image

TL;DR

In twelve months the vendors most mid-market companies run on rebuilt themselves around agents. Salesforce reported Agentforce ARR above $1.5 billion on August 26, 2026, up 240% year over year; HubSpot said its Customer Agent resolved 72% of support tickets without escalation in Q2 2026.

The pricing changed with the products. Agentforce meters agent actions in Flex Credits at roughly ten cents each; HubSpot charges per resolved conversation and per qualified lead. Your CRM bill now moves with what agents do, whether or not anyone designed what they do.

The pressure from above is real and the results are thin. Gartner finds 71% of boards pressing CEOs for an AI strategy and 80% of CEOs expecting AI to force operational overhauls; PwC finds only 12% of CEOs reporting both cost and revenue benefits, and 56% reporting none.

One decision remains with the executive team, and it has nothing to do with which model to buy: whether the company designs how agents enter its operating model, or absorbs the design its vendors ship by default.

The platforms decided first

On August 26, 2026, Salesforce reported second-quarter results for fiscal 2027. Agentforce annual recurring revenue passed $1.5 billion, up 240% from a year earlier. Combined with Data 360 the figure was about $3.9 billion. The company said it had delivered seven billion “agentic work units” to date, 3.2 billion of them in that quarter alone, and that half of its Agentforce bookings came from customers refilling credits they had already used. Two thousand more paying customers went into production during the quarter. In February the count had been 29,000 deals since launch and $800 million of ARR. The line is not flattening.

HubSpot’s August 5 earnings call told the same story at a different altitude. Its Customer Agent resolved 72% of support tickets without a human, roughly 17,000 customers had switched on the Prospecting Agent, more than 16,000 had activated the Data Agent, and 55% of Professional and Enterprise customers were using agents or Breeze Assistant. Monthly agent actions had more than tripled during the year. HubSpot has 306,446 customers. That is a lot of support queues now being answered by software.

Then, on the same day as the Salesforce results, the Claudeforce partnership between Salesforce and Anthropic was announced: Claude inside Agentforce’s reasoning engine, Salesforce data available inside Claude. Whatever one makes of any single announcement, the shape of the year is clear. The vendors that hold your customer record, your pipeline and your support queue have decided that the unit of work inside their products is an agent action, and they have priced accordingly.

The bill moves before the strategy does

That last part deserves a closer look, because it reaches a CFO before it reaches a strategy offsite. Agentforce now meters agent work in Flex Credits: 100,000 credits for $500, a standard action costing 20 credits, which works out to ten cents each, plus $2 for every customer-facing conversation. Enterprise Edition customers received 100,000 credits at no cost through Salesforce Foundations, which is a generous way to start a meter running. HubSpot moved its core agents to outcome pricing in 2026: about fifty cents per conversation the Customer Agent resolves, about a dollar per lead the Prospecting Agent surfaces, drawn from a credit pool that ships with every tier.

Consumption pricing is a rational model for the vendor and a revealing one for the customer. An agent that takes ten actions to close a case that should take three costs three times as much, forever, and the difference between the two is the quality of the data, the permissions and the process the agent was given. The bill is now a readiness score, delivered monthly. Gartner predicted in June 2025 that more than 40% of agentic AI projects would be canceled by the end of 2027, citing cost, unclear value and inadequate risk controls. Consumption pricing makes all three legible on an invoice.

This is the sense in which the world has already changed for a company of 300 to 2,000 people. Nobody at that company chose to run on agents. The choice was made upstream, in the roadmaps of two or three vendors, and it arrives as a feature flag, a credit balance and a line item. The timeline below puts the year’s moves side by side.

What it asks of an executive team

The pressure to respond is coming from above as well as from the stack. Gartner’s 2026 board survey found 71% of boards pressing their CEOs for an AI strategy, with AI the number one investment priority for 57% of directors. EY’s Board Matters analysis found the share of Fortune 100 companies disclosing board oversight of AI risk rose from 16% to 48% in a year. Gartner’s April 2026 CEO survey reported 80% of chief executives expecting AI to force high-to-medium change to their operational capabilities. And PwC’s 29th Global CEO Survey, published in January 2026, found only 12% of 4,454 CEOs saying AI had delivered both cost and revenue benefits, with 56% saying it had produced no significant financial benefit at all. The board is asking. The answer, at most companies, is still a pilot.

What the moment asks of an executive team is smaller than the anxiety suggests, and more uncomfortable. It asks for three things.

Decisions that could be postponed for a decade

Who owns the customer record after the deal closes. What a lead is, in one sentence that marketing and finance both sign. Which system is the truth for the invoice. Who is accountable when an automated decision is wrong. Digital transformation let companies leave those questions open, because a person sat between every system and reconciled the difference by hand, usually in a spreadsheet, usually on a Monday. Agents do not reconcile. They execute, at ten cents an action, on whichever version of the truth they were pointed at. Ambiguity that used to cost a few hours a week becomes a risk with a customer’s name on it.

Gartner’s May 2026 research found that only 27% of executives have a comprehensive AI strategy. Gartner’s CEO survey the year before found 66% saying their business model was not fit for AI. Those two findings describe the same company: one that knows the model has to change and has not yet made the decisions that would change it, because each decision forces a conversation between two departments that have avoided it for years.

A foundation before a feature

Every capability your vendors ship in the next three years lands on the same data, the same integrations and the same permission sets you have today. The 2026 MuleSoft Connectivity Benchmark found 96% of IT leaders saying agent success depends on data integration, and half of deployed agents running in silos. Salesforce’s own CIO Trends 2026 survey found only 23% of CIOs completely confident investing in AI with their current data governance. The companies that benefit from each release without a rebuild are the ones that made their foundation ready once, on purpose: one system of record per object, definitions written down, integrations that carry context, an agent permission model that matches a human in the same role.

This is not the technology project the board imagines. BCG’s rule of thumb puts 10% of AI value in algorithms, 20% in technology and data, and 70% in people and process. The IBM Institute for Business Value’s 2026 CEO study found 83% of CEOs believe AI success depends more on adoption than on the technology itself. The foundation is mostly decisions and hygiene, which is why it gets skipped, and why skipping it is so expensive.

One name

Not a task force. A person whose goals contain the number the agent is supposed to move. IBM’s study reported that 76% of organizations now have a Chief AI Officer, up from 26% a year earlier, and Gartner found 91% of high-maturity AI organizations have dedicated AI leaders. The title is beside the point at 600 people; the COO, the CRO or the head of business systems is usually the right name. What matters is that the outcome sits in one person’s goals and that person can say no to a workflow that wants its own data pipeline. In our experience it is the strongest single predictor of whether an AI workflow reaches production, and it is the cheapest decision on this page.

Protiviti and BoardProspects surveyed directors in late 2025 and found that 26% of boards discuss AI at every meeting. Among organizations reporting high AI returns the figure was 63%; among the rest, 13%. Attention from the top is no substitute for the three things above. It is, on that evidence, closely tied to whether they get done. The numbers below are the executive view of 2026, compressed.

Diagram: Pressure from above, results from few.

What we actually see

A composite from engagements, and inference rather than a controlled study. The executive team has read the board’s question and answered it with a purchase. Agentforce is switched on, or a Breeze agent is live, within a few weeks. It demos well. Three months later the CFO asks about the credit consumption line, the support lead cannot say whether resolution time improved because nobody wrote the number down before go-live, and the sales team has quietly stopped trusting the lead scores because the definition of a qualified lead is still the one marketing uses and not the one finance reports. Nothing failed. Nothing was designed, either. The vendor’s default became the company’s operating model, one credit at a time.

The recovery is unglamorous. The three decisions get made, in a room, in about a week. The agent is re-pointed at the system of record. A baseline is reconstructed from historical data, with an honest caveat attached. The bill goes down, because the agent stops taking ten actions to do three, and the metric becomes defensible. The cost of the detour is usually a quarter and some credibility with the board. We have seen that sequence often enough to expect it, though not so often that we would call it a law.

The decision that is still yours

There is one. It has nothing to do with which model, which vendor or which credit pack. It is whether your organization designs the way agents enter its operating model or absorbs the design its vendors ship by default. The first path is slow to start and cheap to finish: decisions, definitions, one system of record per object, one workflow in production with a number, a governance rhythm, then the next workflow on the same foundation. The second is fast to start and expensive to finish, because it leads to the 56% PwC counted as seeing no financial benefit, and to the 40% of agentic projects Gartner expects to be canceled.

PwC also found that companies with strong foundations were three times as likely to report returns from AI. That multiple is the whole argument. The platforms will keep moving; Salesforce added 2,000 paying production customers in a single quarter and HubSpot tripled monthly agent actions in a year. Waiting for the technology to settle is not a strategy, because the technology is not going to settle and the meter is already running.

Make the three decisions this month: one owner with the metric in their goals, one flagship workflow the board already measures, one system of record per business object, written down.
Run a readiness assessment before the next vendor conversation, and treat every red item as a task with a name attached.
Take one workflow to production, not pilot, with a baseline recorded before go-live and a person who can switch it off.
Read the consumption line on the CRM invoice every month against the outcome it bought. If the ratio is wrong, the fix is in the foundation, not the vendor.

The world is changing. Take it as given that your organization is part of that change; the platforms it runs on made that decision, priced it and shipped it. The open question is on whose terms. That one is still yours, and it will not stay open for long.

Frequently asked questions

What does it mean that CRM platforms are being rebuilt around AI agents?

It means the software already holding your customer, pipeline and support data now acts on it rather than only recording it, and is priced on that action. Salesforce meters Agentforce in Flex Credits per agent action; HubSpot charges per resolved conversation and per qualified lead. The change arrives inside systems you already own, so it applies whether or not your company planned for it.

How does consumption pricing for AI agents affect a mid-sized company?

It turns readiness into a monthly cost. An agent working on clean data with clear permissions and a documented process takes fewer actions to reach an outcome than one pointed at three versions of the customer, and with Agentforce charging roughly ten cents per action the difference compounds. The CRM invoice becomes a signal: if cost per outcome is rising, the foundation, not the vendor, is the problem.

What should a CEO do first about AI?

Make three decisions that cost nothing but a meeting: name one executive whose goals contain the metric AI is meant to move, choose one flagship workflow the board already tracks, and declare one system of record per business object. Those decisions gate everything else. Buying a platform before making them produces a pilot that demos well and cannot be measured.

Is AI transformation different from digital transformation?

Yes, in one specific way. Digital transformation moved work into software and let people reconcile the gaps between systems by hand. AI transformation changes who does the work inside that software, and agents execute rather than reconcile. Questions digital transformation could leave open, such as who owns the customer record and what a lead is, become operational risks the moment an agent acts on the answer.

Should we wait for the AI platforms to mature before acting?

No, because the parts that need your attention do not depend on the platforms. Ownership, definitions, systems of record and permissions will be required by every release for the next several years, and Salesforce and HubSpot are shipping quarterly rather than settling. PwC found companies with strong foundations three times as likely to see returns; the foundation can be built now and the platform chosen later.

Who should be accountable for AI in a 300–2,000-person company?

One named executive with the business outcome in their goals, typically the COO, CRO or head of business systems. IBM reports 76% of organizations now have a Chief AI Officer, but at mid-market scale the title matters less than the mandate: the person must be able to refuse a workflow that wants its own data pipeline and must own the number the board sees.

NEXT STEP

See where your organization stands before the next AI purchase.

A short conversation is enough to tell whether the AI Readiness Health Check, 36 checks on your Salesforce or HubSpot org, is the right first step. No decks, no obligations.

book a strategy session

Read the 36 checks →

WRITTEN BY

Tal Oryon

MORE NOTES

Perspective