Skip to content
PRAXIS

07How we work

Where AI helps this practice, and where it deliberately doesn't.

Praxis runs a dedicated AI advisory practice, so it would be a strange kind of dishonesty to be vague about how AI is used inside its own work. Here's the actual line.

Rows of network server racks in a data centre: an illustrative image for applied AI infrastructure.

Why this line is drawn where it is

Knowing where AI's reasoning breaks came from testing it directly.

Before Praxis existed, the founder worked at the seam of finance and AI directly: analyzing 10-Ks and SEC filings, authoring more than 1,500 valuation and accounting Q&A sets, and building the evaluation frameworks used to test how large language models reason about financial problems. That work is described in full on the about page : it was done at AfterQuery, and it's prior experience, not a Praxis result. What it produced, in practice, is a fairly precise sense of exactly where an AI system's financial and strategic reasoning holds up under scrutiny, and where it quietly breaks. That's the line drawn below.

Where AI genuinely helps

Speed on the mechanical layer.

  1. Market scans, competitor reads, and first-pass literature synthesis that used to take days can take hours.

    That time gets redirected into more scenarios tested and more assumptions pressure-tested: not into finishing the engagement faster for the same money.

  2. 02

    Structuring a messy first draft

    Turning a rough set of notes, interview transcripts, or a client's internal documents into a structured first cut of an analysis: a starting point the advisor then checks, argues with, and rewrites, not a finished product handed over as-is.

  3. 03

    Stress-testing a conclusion before you see it

    Running a recommendation against counterarguments and edge cases before it reaches you, the same way a second reviewer would surfacing the weak points in the reasoning early, while they're still cheap to fix.

Where it deliberately doesn't touch the work

Judgment, not just compliance boilerplate.

  1. Which strategic bet to recommend, when the evidence is genuinely ambiguous and the cost of being wrong is real, stays a human decision made by someone who is accountable for it.

    AI can lay out the trade-offs; it doesn't get to pick which one you should take.

  2. 05

    Anything client-specific and irreversible

    A recommendation that will shape a market-entry decision, an org redesign, or a capital allocation isn't drafted by a model and lightly reviewed: it's reasoned through directly, because the review pass isn't rigorous enough for a decision with real consequences.

  3. No AI system takes the call when a recommendation doesn't hold up, sits in the room for a hard conversation, or is accountable to you for the quality of the work.

    That stays with the advisor, always: see /how-we-work/solo-advisor.

Questions

Before you reach out.

Will I know when AI was used in a deliverable?
If it's relevant to how much weight to put on a specific section, yes: the honesty standard on this site (see /about) covers process, not just claims about results.
Does using AI make engagements faster to deliver?
Often, for the research and drafting layers. It doesn't compress the parts of the work that are genuinely about judgment, which is usually the part that determines how long a good engagement actually takes.
Does Praxis advise other companies on their own AI adoption?
Yes. See /services/ai-implementation-consulting, /services/ai-change-management, and /services/ai-ethics-governance. The same honesty about where AI helps and where it doesn't shapes that advisory work, not just how Praxis runs internally.

More on how we work

The rest of the methodology.

Deciding how AI fits your own business?

That's a different, harder question than most vendors admit. Tell us where you're stuck and we'll give you an honest read.

Start a conversationOr call (415) 634-5434

No obligation · a scoping conversation first