T&M is dead (not really)

Last quarter, SAP's CEO told his earnings call that customers shouldn't be paying their system integrator "$10 more" for migration work AI can handle — that data migration, configuration, and test automation can now be done faster, cheaper, and more reliably without an army of consultants. He isn't alone. Salesforce's CEO has told customers to stop "DIY-ing" their AI and buy the agents directly. Accenture took an $865M charge to exit staff who can't be re-platformed on AI skills. TCS laid off 12,000 in its first major restructuring in decades. The headline everyone is writing is some version of: AI is coming for consulting, and T&M is finally dead.

The headline is wrong about the second part.

The question isn't whether AI will kill time and materials. It won't. But AI is making implementation work more predictable. This changes the pricing conversation entirely and almost nothing is being written about why.

This conversation is older than you think

Every few years someone important declares T&M dead. The early 2010s version was offshoring and packaged accelerators. The mid-2010s version was SaaS — "the implementation is the product now." The late 2010s version was low-code. The 2026 version is agents.

Each wave produced the same prediction, the same panels, and the same result. T&M endured.

It's not because T&M is invincible. It's because every round of this conversation misreads what T&M actually is.

What T&M actually is

Most people who predict the demise of T&M talk about it as a pricing mechanism, but that's too simplistic. T&M isn't just how SIs charge clients; it's deeply rooted in how a firm runs: how consultants are utilized, how margins are measured, how sold-day revenue is forecast. You can't address the pricing model without touching everything built on top of it.

But there's a second dimension that matters more. T&M is, at its core, a risk-shifting mechanism.

When an SI begins an implementation, nobody knows how complex it will become. A six-country SAP rollout scoped at $4M becomes a nine-country program at $11M. A Sales Cloud implementation meant to replace a legacy CRM becomes a reorganization of the client's sales operation. A ServiceNow ITSM project turns out to be 40% about data quality nobody budgeted for. Neither the SI nor the client can predict this at kickoff. T&M passes the risk on to the client: if scope expands, the client pays more.

Every attempt to replace T&M has stumbled on the same question T&M answers: who bears the risk when the work is unpredictable? A fixed fee with a 25% contingency baked in is T&M with a better looking cover sheet.

Under T&M, the SI's interests run against the client's in one specific way: the SI makes more money when the work takes longer. Fixed-fee, outcome-based, and gain-share models all try to flip this. Tie the price to a deliverable, or a result, or a shared metric, and the SI's incentive aligns with finishing well instead of running long.

The reason none of them have scaled is that they all need the same thing: the ability to predict, within tolerable error, what an engagement will cost to deliver. Without this, fixed-fee, outcome-based, and gain-share pricing are all bets SIs can't accurately price. You can't align incentives based on numbers you can't calculate.

What AI actually changes

Most of the coverage of AI in implementation is about efficiency. AI drafts user stories faster. AI generates configuration objects. AI writes test scripts. A task that took ten consultant-days takes one.

While efficiency is real, it doesn't change the pricing model.

Under T&M, efficiency is a problem for the firm. Ten days of billable work becomes one. The client benefits. The SI absorbs the gap. This is why it's structurally hard for partners to champion AI adoption internally. Adoption kills their revenue.

Market competition will push SI project prices down either way. The more relevant shift is predictability.

Predictability lets an SI firm understand what something will cost to deliver before signing a contract. Time taken for requirements capture, configuration, and integration mapping used to vary wildly by engagement, which is what made fixed fees so dangerous. When an agent does that work, variance largely collapses. A firm can confidently price a fixed fee without the 25% contingency that used to blur the line between it and T&M.

Every previous wave of "T&M is dead" tried to change delivery, but the bottleneck was always measurement. Offshoring made hours cheaper, accelerators made hours fewer, SaaS moved some hours out of the picture entirely. None of them changed the unpredictability of software implementation that made T&M necessary in the first place.

Agentic systems are different. They don't just do the work faster. They do it consistently. The second time an agent configures a standard Sales Cloud org, it produces an output similar to the first. Over hundreds of engagements, the variance in what "a Sales Cloud rollout" means can collapse to something a firm can actually price.

Standardization has been the missing precondition for every alternative to T&M that's ever been pitched. Agents are the first technology that delivers it.

Three pricing models, and an idea for another

The SIs getting this right aren't replacing T&M with one thing. They're building fluency across several existing options, and coming up with new ones.

Fixed bid. The most established alternative, now actually defensible for standardized work. The client gets cost certainty. The SI captures the margin compression AI has enabled — which is meaningful, if the firm is honest with itself about its cost base.

Portfolio pricing. A fixed annual fee covering a defined set of recurring work: all Sales Cloud enhancements for a global business, all quarterly SAP releases, all ServiceNow request-management work across twelve months. The client gets predictable spend. The SI gets annuity revenue plus a throughput incentive — the faster the work is delivered, the higher the margin. Portfolio structures also build an embedded relationship that becomes hard to displace. A partner who knows a client's data model, release cadence, and regulatory posture is a different kind of asset than one that shows up for the next RFP.

Outcome and gain-share. Talked about more than practiced. Outcome-based pricing in SI has a mixed record for structural reasons: outcomes are hard to attribute, time horizons are long, and client procurement systems are often literally incapable of processing variable pricing. The firms that make it work do two things. First, they reduce room for debate by selecting outcome metrics the platform itself can actually measure: quote-to-cash cycle time on Revenue Cloud, case deflection on Service Cloud, DSO reduction on S/4HANA. Second, they treat outcome pricing less as economics and more as signaling: a way to tell a buyer, "we are so confident this transformation will work that we'll only get paid if it does."

Pricing the token. The unit of billing in implementation work is shifting underneath everyone's feet. It used to be the consultant-hour. Increasingly, the substantive work is being done by agents. The unit that corresponds to that work isn't an hour. It's an agent-action, a resolved conversation, a token, a completed object.

This unit already exists in every SI's invoice. It's just buried. Watsonx is sold by the token. Agentforce by the conversation or the action. SAP's AI Units are metered compute. SIs pass these through to the client as line items, and then price the judgment layer around them in consultant-hours.

The next generation of SI pricing will look something like this: token costs passed through with a markup, plus a fixed fee for judgment. The firms that let others define it will spend the back half of this decade explaining to procurement why their invoice is denominated in a unit that maps to less and less real work.

The point isn't disruption

The framing of "T&M versus something better" is a false binary. The future of SI pricing isn't one model replacing another — it's a more varied mix, matched to the actual shape of the work. The SIs that thrive will be fluent across all of it.

More predictable delivery means more predictable cost structures. Better cost visibility means more confidence in fixed-fee pricing. The uncertainty that made T&M a rational hedge against complexity diminishes as agents make the underlying work more standardized and repeatable.

T&M will remain where it belongs, on the complex, judgment-heavy engagements where time spent really is the best proxy for value delivered. It stops being the default answer to everything else.

The death of T&M has been predicted in every cycle since the model became the industry's default. Every cycle has been wrong. This one will be wrong too. But the firms still publishing rate cards denominated only in consultant-hours by 2028 will be looking at a very different industry than the one they stand in today.