The Prompt Was Cheap. The Follow-Up Was Expensive.

A prompt can take ten seconds to type and create an hour of work for everyone downstream. "Review this contract for risk" sounds efficient until somebody has to sort through a page of generic observations, find the clauses the tool is talking about, and figure out which issues matter to the client.

We talk about AI pricing in subscriptions and tokens. A firm also pays for the human time it takes to give the tool context, check what comes back, and fix whatever slips through. That's the economics of prompting. What does it cost to get to work you can actually use?

A better prompt can make the answer cheaper to check

Take that contract review. If I want a first pass on indemnity, liability, and renewal, I can ask the tool to identify the relevant provisions, quote the exact language, cite the section numbers, and flag any missing exhibits or cross-references. The lawyer can then go straight to the provisions and see whether the tool found what matters.

That extra minute writing the prompt buys something concrete. It makes verification easier. It also limits the work the tool has to do. If I only need to know whether the agreement auto-renews and when notice is due, a full contract risk memo gives me more reading and more claims to verify.

Reuse changes the math

It may take longer to design a good AI workflow for a single contract than to review the contract yourself. Run the same first-pass check across fifty agreements and the economics start to change. You can reuse the instructions, review the results in a consistent format, and use recurring errors to improve the workflow.

Here's a simple example. Say it takes an hour to build and test a reusable prompt, and it saves six minutes per agreement after verification. You break even at ten agreements. Across fifty, you save four hours after accounting for setup, assuming the same savings hold. If the agreements vary enough that every result takes extra time to check, the savings shrink. The way to find out is to time a small pilot and look at what the tool missed.

Ask who gets the savings

For a fixed-fee matter, a faster first pass can improve the firm's margin or give the firm room to offer a lower price. On an hourly matter, it can reduce what the client pays for a routine step. In either case, track where the time goes. If a junior lawyer saves twenty minutes and a partner spends thirty cleaning up the output, the cost has moved up the billing ladder.

At Delphoria Learning, we teach teams to write prompts around the work they need to finish and the evidence they'll need to check it. Start with a task the team does often, try it on material approved for the tool, and compare the two workflows through completion. Ask the person doing the review what made the answer easier or harder to verify. That feedback tells you what to change in the prompt and whether the workflow is worth repeating.

The meter stops when the work is ready to use. Count from there.

Learn more at delphoria.ai

Next
Next

The Brief, the Bot, and the Fine Print