New opportunity

AI Acceleration Engineer

Decagon · San Francisco; New York City, United States

About this role

Decagon seeks an AI Acceleration Engineer to make AI practical across its business and build the internal systems teams rely on. Based in San Francisco or New York City, this in-office role sits within the AI Acceleration team, currently part of Finance, and works across Finance, Accounting, People, Legal, Sales and Recruiting.

Responsibilities

Own projects from requirements gathering and solution design through development, testing, deployment and ongoing maintenance. Build internal web applications and integrations, extend the contract and customer source-of-truth platform, and replace spreadsheet-based processes with reliable systems.
Automate billing, collections and month-end close work. Address revenue recognition, usage data ingestion and reconciliation among internal data, the ERP and Salesforce.
Identify automation opportunities with business stakeholders and decide whether a custom solution, AI vendor or hybrid approach fits. Build MCP servers, skills and agent-facing tools that let teams work with internal systems through AI, while enabling power users to handle simpler automations.
Measure and report time saved, costs reduced and errors eliminated.

Background:

Requires 6+ years of hands-on experience building internal tools, business systems or automations for business teams, including revenue or finance operations, plus ownership of projects across their full lifecycle.
Must be able to build production-quality tools, reason about data flows and maintainability, choose between agentic AI and deterministic pipelines, and independently lead cross-functional work under ambiguity.
Requires hands-on use of AI-native coding tools such as Claude Code, Codex or Cursor to ship production solutions. Coding in Python, TypeScript or JavaScript, API work and deployment experience are preferred.
Additional advantages include leading AI transformation initiatives; a product mindset; hands-on GCP experience with Cloud Run, BigQuery, Cloud Functions or Pub/Sub; building custom MCP servers; and experience with billing, revenue systems or ERP implementation.

Compensation is $180K–$220K plus equity. Full-time employee benefits include medical, dental and vision coverage, life and disability insurance, a retirement plan, parental and fertility benefits, a monthly wellness and lifestyle stipend, office lunches and snacks, and a take-what-you-need vacation policy, subject to local requirements.

For full details about this role, click “Apply on company site” to view the employer’s original posting.

Skills for this role

AI agentsMCP serversWorkflow automationInternal toolingBusiness systemsWeb applicationsSystems integrationBillingRevenue recognitionData ingestionData reconciliationERPSalesforceClaude CodeCodexCursorPythonTypeScriptJavaScriptAPIsGCPCloud RunBigQueryCloud FunctionsPub/SubRequirements gatheringSolution designStakeholder management

Checking your profile…

Preparation suggestions, not the employer’s actual interview questions. Always check the original posting for current requirements.

Understand the role

What matters most

Own internal systems through maintenance
The role covers requirements, design, development, testing, deployment and ongoing maintenance. It includes extending the contract and revenue platform and replacing spreadsheet processes with dependable tools.
Automate finance and revenue operations
Billing, collections, close, revenue recognition and usage-data ingestion are named workstreams, alongside reconciliation among internal data, the ERP and Salesforce.
Make practical choices about AI
The engineer decides when to use a custom build, an AI vendor or a hybrid approach, and when an agentic workflow is preferable to a deterministic pipeline. The role also builds MCP servers, skills and agent-facing tools.
Lead work across business functions
The role works directly with stakeholders in Finance, Accounting, Legal, People and Recruiting to find use cases and drive projects under ambiguity without formal authority.
Enable adoption and show impact
The posting calls for helping power users handle simpler automations and reporting time saved, costs reduced and errors eliminated to leadership.

Likely challenges to think through

Reconciling data without obscuring discrepancies
Preparation hypothesis: Automating usage, billing and close work across internal data, an ERP and Salesforce may pit speed against accuracy. Prepare to explain how you would trace data lineage, surface mismatches and test results before relying on an automated process.
Choosing AI without adding unnecessary variability
Preparation hypothesis: Agent-facing tools can expand access to internal systems, while some finance workflows may need more predictable behavior. Prepare a decision framework for agentic versus deterministic approaches, including how you would test outputs and constrain system access.
Delivering quickly while retaining ownership
Preparation hypothesis: The stated fast pace and ongoing-maintenance responsibility may create tension between shipping an immediate fix and building a system others can maintain. Prepare an example of how you would scope an initial release, validate it with users and plan its upkeep.
Aligning functions with different needs
Preparation hypothesis: A workflow spanning Finance, Legal, People or Recruiting may have competing definitions of success. Prepare to describe how you would clarify requirements, agree on an owner and measure whether the resulting tool is adopted.

Practice your answers

Technical practice

How would you extend a contract and customer source-of-truth platform while replacing a spreadsheet workflow?

What to coverSuggested preparation: Explain how you would map existing data and decisions, define ownership of each field, and design a migration that preserves trustworthy records. Discuss validation, user access, and how the replacement would remain maintainable.

How would you automate billing, collections, and month-end close work?

What to coverSuggested preparation: Break down the handoffs, exceptions, and approvals that must be understood before automating a finance workflow. Explain how you would test outputs and preserve a way for teams to investigate discrepancies.

How would you ingest usage data for billing or revenue workflows?

What to coverSuggested preparation: Describe checks for completeness, duplicates, timing, and changes to incoming data. Explain how you would trace a disputed figure back to its inputs, while confirming revenue-recognition rules with the responsible accounting stakeholders rather than assuming them.

How would you diagnose a mismatch among internal data, the ERP, and Salesforce?

What to coverSuggested preparation: Compare record identifiers, definitions, update timing, and transformations before deciding which value is authoritative. Outline a repeatable reconciliation process that identifies exceptions and routes them to the appropriate owner.

How would you design an MCP server or agent-facing tool for an internal system?

What to coverSuggested preparation: Describe narrowly defined operations, access boundaries, clear tool outputs, and controls for actions that change business records. Explain how you would test the interface and make failures visible to its users and maintainers.

When would you choose an AI agent instead of a deterministic pipeline?

What to coverSuggested preparation: Compare variability of inputs and decisions against the need for predictable, auditable outputs. Explain where a hybrid workflow might use AI for interpretation but deterministic checks for consequential actions.

How would you decide between a custom build, an AI vendor, and a hybrid solution?

What to coverSuggested preparation: Compare fit to stakeholder needs, integration effort, data handling, ongoing maintenance, and cost. Describe a small evaluation that tests the hardest workflow and defines evidence for a decision.

How would you deliver and maintain a production-quality internal web application or integration?

What to coverSuggested preparation: Walk through requirements, API and data-flow design, testing, deployment, monitoring, and ownership after launch. Discuss trade-offs that help another engineer maintain the system; GCP services are worth mentioning only if relevant to the design, as the posting lists them as an advantage.

How would you use AI-native coding tools to ship a production solution responsibly?

What to coverSuggested preparation: Explain which coding tasks you would delegate to a tool such as Claude Code, Codex, or Cursor and how you would review and test its output. Address how you would protect sensitive business data and retain ownership of the deployed result.

How would you measure whether an internal automation delivered value?

What to coverSuggested preparation: Define a baseline and a method for measuring time saved, costs reduced, and errors eliminated, as requested in the posting. Explain how you would account for maintenance effort and communicate results without overstating them.

Behavioural practice

Tell me about a time you built trust with stakeholders who had different priorities.

What to coverSuggested preparation: Use a real situation involving business stakeholders, explain how you learned their constraints and resolved a disagreement, and state the result. Keep your own actions distinct from the team's work.

Describe a project you independently drove when requirements were ambiguous.

What to coverSuggested preparation: Give the situation, the questions you used to clarify scope, and the decisions you made without relying on formal authority. Close with the delivered result and what changed as you learned more.

Tell me about a system you continued to own after launch when something went wrong.

What to coverSuggested preparation: Set out the operational problem, your diagnosis and response, and how you worked with affected users. Give the result and any change you made to reduce recurrence.

Describe an automation that people initially struggled to adopt.

What to coverSuggested preparation: Explain the users' workflow, how you discovered the barrier, and the changes or support you provided. State the result using observed adoption or workflow evidence rather than assuming launch meant success.

Tell me about a time you shifted between substantially different business problems.

What to coverSuggested preparation: Use a real situation to show how you learned a new domain, prioritized amid competing needs, and delivered a result. Explain what you carried over from the first problem and what required a different approach.

Sharpen your knowledge

Topics to prepare

End-to-end ownership of internal systems
The posting calls for ownership from requirements and design through testing, deployment, and maintenance. Suggested revision: map one internal-tool project across those stages and identify what you would monitor or maintain after launch.
Contract, customer, and revenue data flows
Decagon describes an existing contract-and-customer source of truth and reconciliation across internal data, the ERP, and Salesforce. Suggested revision: draw a data-flow diagram showing which system owns each field and where mismatches could arise.
Billing and close workflows
Billing, collections, month-end close, revenue recognition, and usage-data ingestion are named workstreams. Suggested revision: trace a hypothetical usage record through billing and close, marking questions that would need Accounting’s input rather than assuming a recognition rule.
Choosing the right automation approach
The role requires judgment between agentic AI and deterministic pipelines, and between a custom build, AI vendor, or hybrid. Suggested revision: compare those options for a repeatable reconciliation task and a less structured stakeholder request, including failure modes and maintenance.
Adoption and measurable impact
The posting asks for tools people use, simpler automations power users can handle, and reporting of time saved, cost reduced, and errors eliminated. Suggested revision: define baseline and follow-up measures for replacing one spreadsheet process.

Skills to revise

Build an internal application and integration
Building and maintaining web applications and integrations is explicit; Python, TypeScript, JavaScript, APIs, and deployment are preferred experience. Suggested exercise: implement a small contract-record workflow with a mock API, tests, and a deployment and maintenance plan.
Reconcile records across systems
The role names usage ingestion and reconciliation among internal data, the ERP, and Salesforce. Suggested exercise: create sample records for three systems, match them, produce an exceptions report, and test missing and conflicting values.
Design an agent-facing interface
Building MCP servers, skills, and agent-facing tools is part of the role; prior custom MCP-server experience is listed as an advantage. Suggested exercise: sketch or build a narrowly scoped MCP tool against mock internal records, specifying permitted actions and how failed calls are surfaced.
Use AI-native coding tools with verification
Hands-on use of tools such as Claude Code, Codex, or Cursor to ship production solutions is stated in the background. Suggested exercise: use one to change a small internal-tool workflow, then review the generated code, run tests, and explain what you corrected before deployment.
Turn stakeholder requests into a delivery plan
The role works directly with Finance, Accounting, Legal, People, and Recruiting and independently drives ambiguous projects. Suggested exercise: role-play discovery for a manual collections process, then write a brief scope, acceptance criteria, rollout plan, and impact measures.

Your preparation checklist

Map the role to evidence
Prepare a one-page map of relevant examples against the posting’s 6+ years of internal tooling or business-systems work, full-lifecycle ownership, and production use of AI-native coding tools. Mark preferred experience separately from core expectations.
Rehearse an end-to-end project walkthrough
Choose a real internal tool or automation, if available, and explain the business problem, requirements, design, testing, deployment, maintenance, and what you changed after users began relying on it. Do not present a hypothetical as past work.
Sketch a revenue-systems design
For a hypothetical usage-to-billing workflow, draw data flows between an internal source of truth, an ERP, and Salesforce. Show where reconciliation, error handling, and maintainability would need attention; the posting identifies these as workstreams, not a specified architecture.
Prepare an automation decision example
Create a short comparison of a deterministic pipeline, an agentic approach, an AI vendor, and a hybrid for one business workflow. State the assumptions, failure modes, and reasons you would choose one approach.
Show how you ship AI-enabled tools
Prepare a concrete example of using Claude Code, Codex, or Cursor to ship a production solution, including how you reviewed and tested the output. If relevant, sketch an MCP server or agent-facing interface that lets authorized users work with an internal system; custom MCP experience is an advantage, not a stated requirement.
Quantify adoption and business impact
Bring an example of working with business stakeholders under ambiguity and a simple before-and-after summary of time saved, costs reduced, or errors eliminated. Explain how users adopted the tool and how you measured the result.
Write questions about priorities and ownership
Ask which contract-and-revenue platform changes or billing and close workstreams are most urgent, how the team chooses custom builds versus vendors, and who maintains tools after deployment.

Availability last checked: Nov 4, 2026. Status can change after a check; confirm on the employer’s website. Report an issue · Our checking policy

KEEP EXPLORING

Similar AI jobs

Related skills and specializations in United States. Matched to this role, not your profile.

Explore more jobs

Finding similar opportunities…