Product Manager, AI Workflows @ Atlas

What matters most

• Understand CPA work firsthand

  The role calls for US tax and accounting knowledge, comfort doing junior-accountant-level work, and the ability to turn messy CPA workflows into product features.

• Prototype and evaluate AI quickly

  Atlas expects demoable features within days of discovery sessions, using frontier AI tools and checking model output for domain accuracy.

• Write engineering-ready specifications

  Specs need clear edge cases and testable acceptance criteria; the PM also participates in technical design discussions and understands trade-offs.

• Keep accountants and engineers aligned

  The role involves representing Atlas credibly with both groups while maintaining alignment across a fast-moving roadmap.

Likely challenges to think through

• Speed versus domain accuracy

  Preparation hypothesis: A demo built within days may be persuasive before its tax or accounting output is reliable. Prepare to explain how you would check representative outputs, flag uncertain cases, and distinguish a demonstration from a validated feature.

• Prototype versus shippable scope

  Preparation hypothesis: A working demo may hide edge cases or technical constraints. Prepare to show how you would turn what the demo proves—and what it does not—into scoped specifications and testable acceptance criteria.

• Workflow needs versus engineering trade-offs

  Preparation hypothesis: Accountants may prioritize faithful workflow coverage while engineers face implementation limits. Prepare an example of how you would make the trade-off explicit, choose an initial scope, and keep stakeholders aligned.

Topics to prepare

• US tax and accounting fundamentals

  Revisit a US tax or accounting task you know well; explain its inputs, calculations, checks, and what a junior accountant would do. Relevant because the role calls for 3+ years in the domain and comfort doing junior-accountant work.

• CPA workflow discovery

  Map an accountant workflow from initial information to completed output, noting handoffs and points of friction. This is a preparation exercise, not a specified Atlas workflow; the role translates messy CPA workflow problems into product features.

• Domain-accurate AI output

  List how you would distinguish a plausible AI answer from an accurate one in a familiar accounting scenario, including what must be checked against source information. Relevant because Atlas expects evaluation of AI output for domain accuracy.

• Engineering tradeoffs

  For a proposed workflow feature, compare a narrow first version with a broader one and explain what each leaves unresolved. This is a suggested exercise for the stated expectation to discuss technical design and understand tradeoffs.

Skills to revise

• Perform and explain accounting work

  Work through a familiar junior-level US tax or accounting task, record your checks, and explain the result to a reviewer. The posting explicitly expects comfort doing work comparable to a junior accountant.

• Prompt and assess AI models

  Prompt an AI model on a familiar accounting scenario, then annotate inaccuracies, unsupported claims, and missing information in its response. This practices the stated prompting and domain-quality evaluation work.

• Build a demoable prototype

  Using AI tools you already know, turn one discovered workflow problem into a small working demonstration within a few days. Atlas explicitly expects rapid prototyping; no particular tool is specified.

• Write an engineering-ready specification

  Draft a feature specification for the prototype with inputs, expected behavior, edge cases, and testable acceptance criteria. These are explicit expectations in the posting.

• Communicate across stakeholders

  Practice a short walkthrough of the same feature for an accountant and an engineer, then summarize decisions and open questions. This prepares for the stated need to represent Atlas credibly and keep stakeholders aligned.

• Prepare a US tax or accounting walkthrough

  Choose a real, shareable example of work comparable to a junior accountant’s. Explain the inputs, decisions, checks, and how an accountant would verify the result. The role explicitly calls for this domain fluency.

• Map a CPA workflow problem

  Create a one-page example showing how you would turn an accountant’s discovery-session findings into a defined problem, proposed feature, and open questions. This is a preparation exercise based on the role’s workflow focus.

• Build a small, demoable prototype

  Prepare a working AI workflow demo using nonconfidential example data. Be ready to explain what you built, what it does not do, and how you would scope a version within days of discovery, as the role expects.

• Show how you evaluate AI output

  Bring a prompt, sample outputs, and a short domain-accuracy review. Identify errors or uncertainty, explain your checks, and show how you would revise the prompt or workflow.

• Draft an engineering-ready specification

  Write a brief spec for the demo with its intended behavior, edge cases, and testable acceptance criteria. These are explicit expectations in the posting.

• Rehearse technical tradeoffs and alignment

  Prepare a concise explanation of one prototype decision, its alternatives, and what you would discuss with engineers. Also explain how you would keep accountants and other stakeholders aligned if priorities changed.

• Bring questions for Atlas

  Ask how accountants participate in discovery, what makes a prototype useful, and how the team decides a spec is ready for engineering. These questions target the responsibilities described in the posting.