The 'T&M' trap: why adding tokens to the billable hour isn't AI transformation
The narrative says AI will kill the billable hour. More likely, firms re-engineer it into 'Time + Tokens + Materials' — effort-based billing in modern clothing. Here's the difference that matters when you choose an AI partner, and how to spot it.
Written for CXOs & boards
Every week, LinkedIn fills with confident predictions that artificial intelligence will finally kill the billable hour. The logic is tidy: automation drives efficiency, efficiency destroys time-based billing, and value-based pricing wins the day.
We are not convinced. In fact, we are skeptical.
Right now, the consulting industry's relationship with AI is mostly noise — heavy spending, light value. A large share of the LLM workflows being deployed are cosmetic: a chatbot bolted onto an old process, a demo that impresses a steering committee and changes nothing underneath.
The disruptive pressure on traditional firms is real. But the industry rarely defaults to meaningful change on its own. Given the choice between reinventing a commercial model and finding a clever way to plaster over the cracks of the old one, most firms will choose the plaster. And they are already preparing to do exactly that.
It isn't only the sellers. On the buy side, the value gap is now measurable: in the 2026 Deloitte–HKU AI Adoption Index, 69% of Hong Kong organisations were experimenting with AI, yet only 4% had reached transformational value — a gap we unpacked separately. When so much spend produces so little movement in the business, the pricing model sitting on top of it matters more than ever, because it decides who carries the risk.
The rise of 'Time + Tokens + Materials'
Mark the prediction: the billable hour isn't dying. It is being re-engineered.
Instead of moving to genuine outcome-based partnerships, many established firms will design intellectually elegant, blended pricing models — frameworks that "optimise" across human expert time, AI processing, and infrastructure tokens.
They will call it a "Time + Tokens + Materials" model.
It will look modern. It will feel considered. It will give a client the distinct impression that something structural has shifted.
Underneath the polish, nothing will have. Firms that adopt it will be doing what they have always done: selling generic capabilities rather than solutions, billing for units of effort, and passing the rising cost of renting third-party models straight through to the client. The capability-selling model is stubborn, and its companions — the billable hour and the T&M invoice — will outlive the market's predictions by years.
A patch, not a shift
For an enterprise leader engaging an external partner, this hardens into a simple choice.
Time + Tokens + Materials
- Human expert hours
- AI processing — tokens
- Infrastructure & materials
Billed by units of effort. Cost rises with usage, the value stays unproven, and the risk sits with you.
A fixed, auditable outcome
- One agreed business result
- Measured against a real metric
- A fixed price the partner commits to
Billed for the outcome. Cost is tied to value — and the partner carries the downside.
Two engagement models for enterprise AI: the first re-labels units of effort; the second prices against a business result. ASTRA’s framing.
The superficial patch — a firm that has simply added another "T" to its T&M model, billing you for automated effort and rented infrastructure. Your invoice now carries a token line; your risk profile hasn't changed. If the tool goes live and the business metric doesn't move, you still pay in full.
The performance shift — a firm willing to do the harder, less glamorous work of tying technology to a measurable commercial result, and pricing against that result. Here the partner has real skin in the game.
How to tell them apart
You can usually tell which one you're dealing with in the first conversation. A few questions cut through the veneer:
- What, exactly, am I paying for — effort or a result? If the answer is a rate card (hours, tokens, seats), it's effort.
- What happens if the system goes live but the metric doesn't move? A patch keeps billing. A genuine outcome partner shares the downside.
- Who absorbs the cost of model and token inflation? If it passes straight through to you, you're renting someone else's infrastructure at a markup.
- What is the agreed business metric, and how is it measured? No metric, no outcome — just activity.
- What does "done" mean? "The tool is in production" is not the same as "the number moved."
Where ASTRA stands
At ASTRA, we won't play the "Time + Tokens + Materials" game. An AI tool going "live" is irrelevant to us if the underlying business metric hasn't moved.
We don't sell hours, and we don't bill for token counts. We partner with institutional leaders to close the gap between technical hype and operational reality — the unglamorous groundwork of data alignment, process governance and workflow architecture — and commit to fixed, auditable business outcomes. It is the same reason we run what we build rather than hand over a slide deck, and embed senior engineers next to your team instead of billing a pyramid.
The industry has a choice to make: cosmetic pricing patches, or genuine structural change. We know which side of the line we stand on.