Engagements
How pricing works
We do not sell time by the day. Billing by time rewards taking longer, which is the opposite of your interest.
Why we do not bill by the day
Day rates make sense when the requirement is fully defined and the work is standardised. Forward deployed engineering is the opposite: the requirement is usually unclear and has to be defined on site. Charging by the day in that situation creates two problems. First, the time we spend understanding your business is billed to you. Second, nothing in the arrangement rewards us for finishing sooner.
More importantly, a day rate means you are buying an engineer's hours rather than an outcome — and when the outcome is poor, responsibility becomes ambiguous.
So our unit of pricing is a phase with a defined scope and defined criteria, not a person-day.
Three pricing models
Which one applies depends on how well-defined your requirement is, and at which stage you want to carry the uncertainty.
Model 1
Fixed scope
Requirement is reasonably clear
One scope, one total price, agreed up front. Suits a bounded pilot or a standalone delivery: what is in, what is out, and what counts as done, all written down before work starts.
- Pricing
- Fixed total, fixed scope
- Boundary
- Anything outside the scope becomes the next phase rather than being absorbed into this one
Model 2
Staged acceptance
Requirement becomes clear as work proceeds
Quoted and settled by phase, each with its own outputs and acceptance criteria. There are stop points between phases: if one does not meet its criteria, the next does not begin. Suits work where direction needs validating as you go.
- Pricing
- Priced by phase, settled in instalments
- Boundary
- You can stop at any phase rather than paying for a direction that no longer holds
Model 3
Outcome and retainer
Long-term, several use cases
A fixed commitment per month or quarter against agreed business indicators. Fits teams that have validated the approach once and now want to deploy it systematically.
- Pricing
- Fixed periodic fee plus agreed indicators
- Boundary
- The indicators must be defined up front, or the model becomes awkward for both sides
Five things that drive the price
The same sentence — "build us an AI support agent" — can vary several-fold in cost. The variation comes from these five factors, not from our pricing strategy.
How many existing systems, and how open they are
How many systems need integrating, whether they expose APIs, whether the documentation is complete, whether the vendor cooperates. A closed interface or a vendor approval requirement can multiply the integration work.
How usable the data is
Where it lives, whether it is clean, whether it is labelled, whether it can be extracted at all. This is the most common source of hidden cost, and the first thing we check during assessment.
How many use cases, and whether they relate
One standalone use case and three interdependent ones are not a factor of three. Related use cases share underlying components and cost less per unit; independent ones are priced separately.
How strict acceptance is
"It works" and "accuracy no lower than 95%, reproducibly" are entirely different amounts of work. Stricter criteria demand more evaluation data, tuning and regression testing.
How well your side is set up
Whether a domain expert is named, whether the decision maker is engaged, whether data and permission approvals move. This factor affects schedule more often than technical difficulty does.
From first contact to a quote
Quotes are not guessed. A number produced without enough information is a risk to both sides.
- 1Step 1
Conversation
We discuss your situation and what you are trying to solve, and decide whether an assessment is warranted. No charge.
OutputA decision on whether to continue
- 2Step 2
Assessment
We come on site, look at the process, the data and the systems, and turn a vague requirement into an executable scope. This is the smallest billable unit.
OutputWritten assessment, draft scope and acceptance criteria
- 3Step 3
Quote
Based on the assessment, we quote either a fixed scope or a phase, stating what is included, what is excluded, and what counts as done.
OutputWritten proposal and pricing
- 4Step 4
Start
Work begins only once both sides confirm scope and acceptance criteria. If confirmation changes the scope, we return to step three.
OutputContract and project plan
An honest note on market rates
We do not publish our own prices, because the cost of the same sentence varies too widely. But some market context is useful so that you are not negotiating without a frame of reference.
Note: the figures below are ranges discussed publicly in the industry and drawn from third-party research. They are not our prices and do not represent our pricing.
- Independent FDE-style projects in China generally run from several hundred thousand to several million RMB, depending on complexity and duration
- Measured as headcount, the ceiling for an FDE role in China is roughly RMB 30,000–40,000 per person-month; above that, the engagement is usually no longer standard staffing
- Some vendors quote embedded staff by the day (public figures include RMB 3,500–5,000 per person-day) or as an annual on-site package
- US FDE salaries span a very wide range, and the same job title can differ several-fold between employers — evidence that the title itself is not a pricing basis; what the person actually does is
Two pricing patterns to be wary of
An FDE project quoted by the person-day
If a firm prices FDE work as "day rate × number of days", it is doing staff augmentation. That is not necessarily wrong, but be clear about what you are buying: hours, not an outcome, with the project risk sitting on your side.
A quote clearly below cost
The cost floor for AI deployment work is estimable: somebody has to understand your business, handle your data and integrate your systems. A quote well below that floor usually means one of three things — part of the scope is hidden, quality will be compressed, or there will be follow-on charges. Ask what is included and what is not.
A quote with no acceptance criteria
A price given without knowing what counts as finished cannot function as a commitment. These engagements tend to end in dispute: you believe it is incomplete, the other side believes it is done.
Common questions
Can you give a rough price up front?
We can give a range, but it is of limited use — cost for the same category of work often varies several-fold, so the range is too wide to decide on. The more useful route is a short assessment that narrows the scope, after which a quote can actually be committed to. The assessment is the smallest billable unit.
What does an assessment cost?
Assessments are billable; the amount depends on how much on-site investigation and how much time is involved. We state the fee and the deliverables during the initial conversation, and only begin once you confirm. If the conclusion is that you should not proceed, we write that down — the report still has value to you.
What if the project stops midway — what happens to what we have paid?
We settle for completed phases. Because we work on staged acceptance, each phase has its own outputs; wherever it stops, billing stops there. We do not ask you to pay for work not delivered.
What is the payment schedule?
Payment follows phase milestones, with proportions set in the contract. We do not ask for a large upfront payment — demanding full or majority payment in advance simply transfers project risk to you.
Do you take small projects?
If the use case is focused enough, yes — the fixed-scope model suits small validation work. What we do not take on is "very small budget with a vague scope", because those engagements almost always end with both sides dissatisfied.
Want a price we can actually commit to?
Start with the assessment. A quote given after the scope is narrowed is one we can stand behind — a number produced without enough information is a risk to both sides.