Price above cost to serve
Every plan is priced above its blended cost per answer, including search, agents and connectors. Free is capped per day so it is a sample, not a subsidy.
Research · Economic
The economics of an assistant, role by role.
Tokens are a cost, not a value. We value an assistant by how much faster a defined role reaches a defined outcome, and we price the product so that value is shared with the customer and the company stays profitable from the first subscription.
Method
Time-to-outcome: a two-week study any company can run
The framework behind ELP GPT Business pilots. It needs no data science team and no access to message content.
- Define the role and the outcome. An outcome is a unit of finished work: a first-draft proposal, a resolved support ticket, a graded assignment with feedback, a reviewed pull request.
- Baseline for one week. Staff log start and finish times for each outcome using their current tools. Ten to thirty outcomes per person is enough.
- Assist for one week. The same staff, the same outcome types, with ElpGPT Business and the relevant plugin enabled.
- Compare medians, not means. Report the change in median time-to-outcome, the share of outcomes that needed rework, and staff-reported confidence. Means are dominated by the hardest cases.
- Translate to money. Value per month = outcomes per month × minutes saved × loaded cost per minute. Compare against the plan price. The pilot should pay for itself inside the second week or it is not a fit.
The independent literature points the same way. Field studies published in 2023 found productivity gains of roughly 14% for customer-support agents with an assistant (Brynjolfsson, Li and Raymond), around 40% faster professional writing tasks (Noy and Zhang, Science), and a 12% increase in task completion with 25% faster completion for consultants on tasks inside the model's competence (Dell'Acqua et al., Harvard Business School working paper). The same consultant study found performance fell on tasks outside that frontier, which is why our plugins narrow the assistant to the work it is good at.
Findings
Where the value concentrates
Patterns from our own pilots and from published studies. Figures are ranges, not guarantees; every organisation should run the two-week study.
| Role | Outcome | Typical change in time-to-outcome | What drives it |
|---|---|---|---|
| Research and analysis | Sourced brief on a market or competitor | 35–55% faster | Live search, source handling, structured summaries |
| Customer support | Resolved ticket, first reply | 15–30% faster; larger for newer staff | Drafting, policy recall, tone |
| Sales development | Qualified first meeting booked | 20–40% more outreach per hour | Account research, personalised drafts, follow-up discipline |
| Education | Lesson plan with differentiated materials | 30–50% faster | Explanations at several levels, question banks, rubrics |
| Software development | Reviewed change merged | 10–30% faster on review and debugging | Codebase explanation, diagnosis, test drafting |
| Operations | Procedure documented and followed | 25–45% faster | Runbooks, checklists, connected-tool actions with approval |
Unit economics
Why ELP GPT is profitable from the first customer
The CMO's version is simple: we do not buy a brand, we build a model that returns money. Here is the operating version.
Routing is the margin engine
Cost policies route each turn to the cheapest provider that meets the quality bar for that task. As providers compete, our cost falls and our price holds.
Acquisition paid in product
Referral credits cost us messages we already serve at marginal cost, not cash. Share cards and link previews acquire customers for the price of rendering an image.
No growth-at-a-loss
We do not run paid acquisition at negative contribution. Growth comes from use, referrals and business accounts that passed their own two-week study.
Capital for capacity, not burn
Proposed infrastructure investment in Ontario is for owned capacity that lowers cost to serve. It is financed against demand, not spent ahead of it.
Investor returns from operations
Returns to shareholders come from operating cash flow. That is a different company from one that returns capital only on an exit.
Statements about pricing and routing describe how the product is operated. Financial figures published elsewhere on this site are company-supplied, forward-looking where marked, and not independently audited. Nothing on this page is investment advice or an offer of securities.
Run the two-week study with us
We will help you define outcomes, set up the pilot and read the results.