OSASI.ORGCommentary · Ideas for our profession
All briefs← Back to OSASI

Prompt engineering is dead. Domain expertise and deterministic models are not

October 20, 2026 · OSASI Commentary

Imran Ahmad, data scientist and author of 50 Algorithms Every Programmer Should Know, says prompt engineering is dead. His bigger point is that algorithms are not. When a bank declines a mortgage or someone prices a house, the answer has to be reliable and explainable, and an LLM is probabilistic and not a deterministic system.

How a brief is made

A member reads an article or listens to a podcast and sends the snippets, key thoughts or concerns tied to a passage or a timestamp. An AI assistant, chosen for the task and the setup (Claude, ChatGPT, Codex, Gemini or another), searches the full article or transcript, pins the times, checks the quotations, and assembles a draft. The draft goes through rounds of review, and nothing is published until a person approves that exact version. The cues on each post start the player at the passage discussed. Commentary is OSASI's perspective, offered for discussion, not a summary of the source. Each brief is co-authored by a person and an AI assistant. AI assistant for this brief: Claude (Anthropic).

Source podcast · Value Driven Data Science

This episode prompted the commentary below. It is one perspective, offered for discussion.

Value Driven Data Science · October 15, 2026
Episode 127: From Algorithms to Agents and What It Means for Data Science

Listening highlights · Dr Genevieve Hayes, host, with Dr Imran Ahmad. Times from the show's transcript, set to Spotify's player (about 4 seconds later).

−13:08 to −12:23 remaining"Prompt engineering is dead"But algorithms are still very important
−12:23 to −10:50 remainingMortgages, house prices and probabilityWhere "we do not want a probabilistic model working here"
−8:21 to −7:27 remainingThe declined mortgageCan the bank show the refusal was not about race or religion?
−3:31 to −2:29 remainingGenAI is not the silver bulletA simple algorithm replaces a frontier model, with better results

Tap a time above to start the player there. If Spotify shows a card about the show first, close it and the player carries on at that time. Times are approximate transcript times; ads and episode editions may shift them. This plays the full episode, not an extracted clip.

Broad AI and deep AI

Ahmad splits AI in two. Broad AI, like an LLM, is a "jack of all, master of none". Deep AI is an algorithm built to do one job as well as possible. His examples of jobs that need the deep kind: a doctor separating benign from malignant tumours, a bank manager approving a mortgage, and "a real estate agent want to predict that how much will be the price of this house." For those, "we do not want a probabilistic model working here." An LLM, "doesn't matter how you describe it, it's a probabilistic system."

The declined mortgage

His sharpest example is a refused mortgage. If the applicant goes to court saying the refusal was due to race or religion, "can bank defend that this was not due to race or religion?" The bank needs "that transparency, that lineage with each of the decision. LLM does not give you that." In healthcare, finance and government, he adds, that transparency is often required by regulation.

The underlying consistency problem in appraisal

In 2021, following a series of racial bias cases which gained widespread media attention, a CBC Marketplace hidden-camera test had one house appraised six times, with White, South Asian and Black occupants. The average for minority occupants came in about 2 percent below the average for White occupants, which six reports cannot settle either way. The bigger finding went unreported: the six values for the same house ran from $1,700,000 to $2,050,000, a spread of $350,000, or about 18 percent of the $1,890,000 average. That was simply swept under the rug.

What this means for OSASI

Appraisal sits in the same place as the bank. A value opinion has to be reproducible and explainable to a lender, a reviewer or a court, and the first defence against a bias claim is a method that gives the same answer for the same house and shows why. That is a deep problem: the appraiser's domain expertise chooses the method, and a deterministic model gives the same answer twice and shows its work.

Ahmad swapped a frontier LLM for a simple purpose-built algorithm and cut the cost from thousands of dollars to zero, with better results. A concept that OSASI calls out in the EOI.

Send a thought →

A like is counted once per reader per day and stores nothing about you. A thought goes by email to the person who edits these briefs; nothing you write appears on the site unless it becomes part of an approved brief. Share opens the service you pick; this site adds no trackers.

Prefer a feed reader, Slack or Teams? Follow by RSS.