Senior Product Manager - Intelligence
Confirmed on the employer's own hiring board on Oct 14, 2026. First seen by Alion on Oct 12, 2026.
We're Osome - an international fintech startup making the lives of entrepreneurs easier. We help thousands of businesses kick admin, accounting and bookkeeping out of their day-to-day, so they can spend more time on what's important to them. We've developed a unique solution that combines SaaS with a human-in-the-loop approach to provide full-fledged services in real-time.
We’re experiencing rapid growth in both client base and team, which now comprises around 500 people across our offices worldwide. As we continue to scale, we’re looking for more bright minds who are excited to make a real impact by solving complex problems.
What to expect:
Intelligence builds the AI capabilities no single team at Osome should carry alone, and helps every other team ship AI well. It works in three modes. It owns cross-cutting capability. It enables product teams by pairing with their engineers, so the pattern transfers and the team runs it afterwards. It incubates only by exception: building inside another team's area when that team has no capacity, always with a written hand-back plan.
This is a hands-on PM role. You'll find problems by talking to end users, dig into the data in BigQuery and Postgres to prioritise, prototype cheaply before committing engineering effort, and ship small fixes yourself. You'll also be the team's main liaison with Operations.
At a glance
You'd own: client and agent assistants (e.g. Maria, our client-facing bot, and the workflows it runs), document and accounting intelligence (e.g. document recognition, AI accounts review, our XBRL agent), and the shared AI foundation (evaluation and safety, memory, Osome MCP and model choices).
You'd work with: five engineers, plus shared design and an analyst. Product teams across Osome and our operations teams are your customers.
The problem to solve: AI is spreading across our teams faster than shared standards are. Your job is to stop every team reinventing evaluation, safety and memory, while our core AI products keep improving quarter on quarter.
What you'll do:
Own the Intelligence roadmap, OKRs and delivery, with two or three big bets each quarter. You make the prioritisation calls, own the business outcome behind the backlog, and write the release notes when things ship.
Set the evaluation bar for AI across Osome: eval datasets, confidence thresholds, human-review sampling and go/no-go gates.
With our Security team, set the safety and data-handling bar for AI across Osome: escalation rules, pre-action checks on agents, fail-open vs fail-closed behaviour, privacy and retention, client data isolation, and provider governance (e.g. zero data retention agreements).
Define the success metrics for each AI product before it ships, then keep checking the data yourself after launch to see whether it's working. Brief the analyst on what to build, and review the result. Use that evidence to make model and provider calls, e.g. benchmark a new model against our own labelled documents before switching.
Break epics into clear stories and tasks that engineers can pick up without a meeting. Pair with product teams' engineers to build AI features, and define what "successfully handed over" means.
Be the team's liaison with Operations: run training and feedback rounds, collect baseline data, and drive adoption until the AI version beats the manual process.
Run a small research track: test new techniques and approaches (e.g. retrieval, memory or agent patterns) on our own problems, and write up what we should adopt and what we should skip.
Represent this team's priorities to senior stakeholders.
As the team grows, mentor other PMs and help manage them.
Who you are:
Evaluated against our PM capability framework. At the Senior PM level, that means:
Strategic impact: you can define and drive a strategy spanning 12 months or more for a platform-style team whose customers are other teams, tied to a company-level goal such as gross margin.
Domain ownership: you own the OKRs, roadmap and business outcome of a complex, ambiguous, cross-functional area. You understand applied LLM/ML products deeply (evaluation, retrieval, memory, agent safety), and can become credible in our accounting and operations domain within two quarters.
Collaboration: you drive high adoption of new tools with operations teams, including the processes needed to run them. You build consensus among senior leaders with conflicting goals, and you can say no while leaving the stakeholder feeling heard.
Problem-solving: you map ambiguous, system-level problems, and know when a rule, a classifier or an LLM is the right tool.
Execution: you reliably ship complex, cross-functional launches involving Operations and other product teams. You prototype to de-risk before committing engineering time, and you spot risks early.
AI-literacy: hands-on well beyond prompting. You build prototypes and ship small PRs with AI coding tools, build eval datasets and run model experiments, read traces, and judge where models fail.
Leadership: you shape direction across teams you don't manage, improve how the team works, and are ready to mentor at least one other PM as a player-coach.
What you bring
AI/ML features shipped to production, with quality measured after launch (accuracy, coverage, deflection).
Experience building evaluation datasets and running model experiments: thresholds, go/no-go gates, human-in-the-loop decisions, and reading traces in an observability tool (we use Langfuse) to see where models fail.
Experience shipping AI features that handle sensitive or financial data, and deciding when they should escalate to a human or stop.
Daily use of AI coding tools (e.g. Claude Code, Codex, Cursor) to read code and ship small changes.
Strong SQL. You can pick the right metrics for a product and explain what you need to an analyst. Data modelling is not required.
Experience working directly with operations teams: running training, collecting feedback, and proving time saved against a manual baseline.
A track record of good calls under ambiguity, and of checking afterwards whether they were right.
(Nice to have) Setting company-wide safety and data-handling rules for AI: fail-open vs fail-closed policies, privacy and retention, and client data isolation, including knowing when a new AI vendor needs a security review before it sees any data.
(Nice to have) Bookkeeping or accounting literacy: enough double-entry, chart-of-accounts and GST/VAT knowledge to judge whether an AI accounts review or document check is right. Either a qualification or hands-on bookkeeping experience counts.
(Nice to have) Document AI/OCR in production: classifying or extracting invoices, bank statements and receipts, and the trade-offs between coverage, accuracy, confidence thresholds and cost.
(Nice to have) Agentic systems or MCP in production: agents or tools that take real actions, with permissions, human approval steps and multi-step evaluation.
Our Benefits
Multi-cultural environment to enhance your global perspective.
Learning opportunities and mentorship from peers and leaders.
Flexible working arrangement; Hybrid working arrangement
An agile working model focused on goals and performance.
Equal Opportunity Statement
At OSOME, creating a culture where individuals of all backgrounds feel comfortable really matters.
Everyone who applies will receive fair consideration for employment. We do not discriminate based upon race, colour, religion, sex, sexual orientation, age, marital status, gender identity, national origin, disability, or any other applicable legally protected characteristics in the location in which the candidate is applying. We want to ensure that we represent the diversity of talent in the society we live in today.
If you have any accessibility requirements that would make you more comfortable during the application and interview process, please let us know so that we can support you.