Data. Intelligence. Results.

Three words, in that order — and it's a process, not a slogan. Take in the messy reality of an operation, find what actually matters inside it, and hand back something you can act on Monday. The domain changes: a public website facing a compliance deadline, a procurement portal nobody has time to watch, a pile of spreadsheets that disagree with each other. The process doesn't.

How we work

One process, three stages

Every engagement runs the same way, whatever the industry. It's the reason the work is checkable — you can see which stage a number came from.

Data — take in the messy reality

Your WMS export, a federal solicitation feed, four spreadsheets nobody trusts. We start with what you already have, however unusable it looks, and consolidate, clean, and validate it into something that can actually be reasoned over. Nothing gets invented to fill a gap — missing is recorded as missing.

Intelligence — find what actually matters

The analysis layer. Where there is a right answer, it gets computed rather than guessed. Where it takes judgment, it is labeled as judgment with the reasoning attached. This is the stage that turns a dataset into a dollar figure, a shortlist, or a ranked set of moves.

Results — something you can act on Monday

A deliverable a person can use without a manual: the twenty changes worth making, the three opportunities worth pursuing, the report that now writes itself every week. Every number traces back to your data, so any of it can be checked.

What we do

The same three stages, pointed at different problems

These are applications of the process, not separate products. If your problem isn't listed, it's probably still the same three stages.

Accessibility & digital compliance

WCAG 2.1 AA audits, conformance reporting, and VPAT/ACR preparation for public entities working toward the ADA Title II deadlines — April 2027 for populations over 50,000, April 2028 for everyone under. Automated scanners catch a fraction of the issues; the human verification is the deliverable.

Public-sector bid response

State, county, city and school-district solicitations: reading the whole package, judging honestly whether it fits, and drafting the response against what the evaluation criteria actually ask for. A person decides what gets submitted — always.

Document compliance operations

Safety data sheet libraries brought to the HazCom 2024 structure and kept current, GHS labels generated from the sheets you already have, and FDA nutrition and allergen panels calculated by the database method rather than sent to a lab.

Data operations, read by an operator

Scattered files and exports merged into one master dataset you can trust, then the repetitive analysis set to run on its own. Written by someone who spent a career generating this kind of data on a warehouse and shipping floor — so the errors that get found are the ones that actually cost money.

Built and running

Pipelines, not slideware

Systems that exist and run today. Each is labeled with what it actually is — in production, or in demonstration. No client names, no borrowed logos.

Built · paused

Federal opportunity sweep

An automated pipeline that pulls federal solicitation and grant feeds, filters them against a capability profile, and produces a qualified shortlist with a bid / no-bid read. It flags what it isn’t sure about rather than guessing. It ran nightly and unattended; the schedule is paused while attention is on direct client work, and it runs on demand.

Data → scraped feeds · Intelligence → fit scoring · Results → a morning shortlist

In demonstration

Pick-path optimization for a warehouse floor

Built from direct experience running pick operations: it models a real floor’s layout and racking, then computes the provably shortest pick sequence — checked against brute force, not estimated. It quantifies the travel waste in the current sequence in dollars.

Data → order & slotting history · Intelligence → exact route math · Results → a ranked moves list

In production

The system that runs this firm

99 Strategic runs on its own tooling: a multi-agent operations system with an append-only decision ledger, refusal logging when an agent lacks grounds to answer, and a human approval gate on anything that ships, sends, or spends. We build the automation we sell.

Data → operational ledgers · Intelligence → agent analysis · Results → work that ships

What makes the output trustworthy

The same rules apply at every stage — they're the reason a number from us can be checked instead of taken on faith.

Exact where it must be exact

Anything with a right answer is computed, not guessed. Deterministic beats plausible.

Honest when it does not know

Absent facts are flagged, never filled with a confident guess. “I don’t know” is a real answer.

Everything traced

Numbers carry their source and every claim can be checked back to the data behind it.

A human signs off

Nothing ships, sends, or spends without a person approving it. Automation assists; it does not decide.

Fixed-scope, reviewed with you

A clear quote up front and work reviewed with you before anything is final — no open-ended meters.

Your tools, your data stays yours

Built around the systems you already use; confidential handling and working copies deleted on request.

Who's behind this

One operator who builds the tools

99 Strategic is a founder-led practice run by Tuan Le — a career spent running operations and working warehouse and logistics floors, now building the data and automation systems himself. You work directly with the person doing the work.

Tell us the problem in plain words

The operation that never quite runs smooth, the messy dataset, the decision nobody has time to make properly. If the process fits, we'll tell you how. If it doesn't, we'll tell you that instead.