Problems we solveCommercial decisions are taken, but rarely recorded or verified.
FBED builds and runs the infrastructure that records each decision, verifies its outcome and learns from the difference. These are the problems it is built to solve, and where each one stands today.
01
Pipeline is unpredictable.
Operating
Why it persists
Most commercial decisions are never recorded with the options that were available, so nothing about them can be measured or improved.
What FBED does
Every decision is logged with its state and the probability given to every option, before anything is sent, then followed to its outcome.
02
Agents reason from raw text.
In development
Why it persists
Reading transcripts is slow, costly and inconsistent, and two reads of the same call rarely agree.
What FBED does
Nyquist parses communication into typed rows; Huffman scores communication as signal and as behaviour, and returns the edits needed.
03
No one can prove what caused a deal.
Measuring
Why it persists
Attribution in sales is assumed, not measured, so spend follows opinion.
What FBED does
Decision graphs with causal uplift, measured against a randomised holdout.
04
Commercial AI has no verified reward.
Task sets from November 2026
Why it persists
Agents are trained on text that looks right, not on what actually happened next.
What FBED does
Task sets across about 25 decision points, each labelled by real, observed outcomes.
05
Humans and machines need to be understood.
Our long-term aim
Why it persists
As machines take on more decisions, accountability for them becomes harder to trace.
What FBED does
Decisions and tools that support human performance, with every action and its reason kept on record.
Meet the five models behind this work.