New opportunity

Senior Associate/Assistant Vice President, AI Product Manager

Temasek · Singapore

About this role

The AI Product Manager owns the end-to-end lifecycle of one or more internal AI products, from identifying problems and building business cases through production deployment, adoption and continuous improvement. The role focuses on investment, portfolio monitoring, research and operations workflows, including products and automations built using an agentic AI platform.

Responsibilities

Define product vision, roadmaps, priorities and measurable business outcomes. Research users and workflows, assess opportunities, and turn findings into product backlogs and delivery milestones.
Make informed AI product design decisions, including RAG versus fine-tuning or prompt engineering, agent workflow reliability, human-in-the-loop checkpoints and evaluation methods. Define success metrics, evaluation datasets and benchmarks across model and prompt versions.
Translate business needs into actionable specifications covering data, model capabilities, integrations, acceptable error levels, system prompts, agent behaviour, tools, API contracts, guardrails and user feedback. Connect usage, accuracy and latency metrics to time saved, decision quality and risk reduction.
Lead delivery with AI engineers, data engineers, UX designers and domain specialists. Work with security and governance, change management and platform engineering teams on compliant delivery, adoption programmes and reusable platform capabilities.
Enable non-technical employees to build approved low-code, no-code and AI-assisted solutions. Develop role-specific learning pathways, starter templates, prompt patterns and agent workflow blueprints; run workshops, office hours and team sessions.

Requirements

4–8 years of product management experience, including at least 2 years owning AI or ML products from problem framing through production deployment. Candidates should have shipped adopted production AI products with measurable impact, not only prototypes.
Technical depth to discuss prompt architecture, RAG design, model selection and evaluation frameworks with engineers. Experience with LLM evaluations, including automated evaluations, human review and red-teaming; strong problem framing, stakeholder communication, outcome measurement and facilitation skills.
Hands-on experience building and deploying workflows with Microsoft Power Platform, Copilot Studio, Anthropic Claude or equivalent agentic AI tools.

Advantages: Experience in investment research, portfolio management, financial operations or risk; experience in a fast-moving AI-native organisation or a financial institution with a substantial AI product function; and familiarity with prompt engineering and agent configuration. The position is listed as permanent in Temasek’s Technology department in SG, 238891.

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

Skills for this role

AI product managementProduct discoveryProduct strategyUser researchProduct roadmappingAgentic AIAgent workflowsRetrieval-augmented generation (RAG)Fine-tuningPrompt engineeringLLM evaluationEvaluation dataset designHuman-in-the-loop designAI governanceStakeholder managementChange managementMicrosoft Power PlatformCopilot StudioAnthropic ClaudeLow-code developmentNo-code developmentWorkshop facilitation

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 AI products through measurable adoption
Source-backed: The role owns internal AI products from problem framing and business cases through launch, adoption tracking and improvement. Prepare an example that connects a production release to sustained use and business outcomes, not just deployment.
Discover and prioritise investment workflows
Source-backed: The manager researches users, shadows workflows and turns pain points across investment, portfolio monitoring, research and operations into sequenced roadmaps. Prepare to explain how workflow evidence shapes scope and priorities.
Make and measure AI design decisions
Source-backed: Responsibilities include choosing among RAG, fine-tuning and prompt engineering; designing agent reliability and human checkpoints; and maintaining evaluation datasets and benchmarks across versions. Prepare to explain decision criteria and how performance changes would be detected.
Translate needs into deliverable, governed products
Source-backed: The role specifies data, integrations, acceptable uncertainty, agent behaviour, tools and guardrails while coordinating engineers, domain specialists, security and governance. Prepare to show how a business need becomes a specification teams can build and review.
Drive adoption and citizen-developer enablement
Source-backed: The manager partners on rollout and feedback loops, develops approved starter templates and learning pathways, and runs hands-on sessions for non-technical employees. Prepare to distinguish production adoption from workshop attendance or initial activation.

Likely challenges to think through

Balancing useful automation with dependable outputs
Likely challenge, not a confirmed employer problem: More autonomous agent workflows may save time but increase error and oversight risk. Prepare a scenario showing how you would set error tolerances, place human checkpoints and use automated and human evaluations to decide what can be released.
Proving value when workflow benefits are hard to isolate
Likely challenge, not a confirmed employer problem: Usage, accuracy and latency do not alone establish better decisions or reduced risk. Prepare a measurement approach that links product indicators to time saved, decision quality or risk reduction, while stating what the evidence cannot prove.
Scaling employee-built solutions without fragmenting controls
Likely challenge, not a confirmed employer problem: Broad low-code access can accelerate useful solutions while creating inconsistent patterns across teams. Prepare to discuss approved templates, role-specific training, feedback and when a recurring need should become a reusable platform capability.

Practice your answers

Technical practice

How would you discover and prioritise an AI opportunity in an investment or operations workflow?

What to coverSuggested preparation: Describe how you would research users, shadow the workflow and distinguish a recurring pain point from an attractive demo. Explain how opportunity size, feasibility and measurable outcomes would shape a scoped backlog and roadmap.

When would you choose prompt engineering, RAG or fine-tuning for an internal AI product?

What to coverSuggested preparation: Compare the options against the task, available data, required quality and ongoing maintenance. Explain what evidence you would seek from engineers and evaluations before committing to a design.

How would you design a reliable agent workflow with human-in-the-loop checkpoints?

What to coverSuggested preparation: Identify where tool use, uncertain outputs or consequential actions warrant constraints or human review. Discuss how you would specify agent behaviour and test failure paths rather than assuming a successful demonstration is sufficient.

How would you evaluate whether an AI product improves across model and prompt versions?

What to coverSuggested preparation: Cover representative evaluation datasets, task-specific rubrics, automated checks, human review and red-teaming. Explain how benchmarks and error trends would inform release decisions and ongoing improvement.

What would you include in a product specification for an AI workflow?

What to coverSuggested preparation: Address data and model capabilities, integrations, acceptable error or uncertainty levels, system prompts, agent behaviour, tools and API contracts. Include guardrails, evaluation criteria and a way to collect user feedback so engineering and business teams share testable expectations.

How would you connect AI product performance to a business case?

What to coverSuggested preparation: Distinguish usage, accuracy and latency from outcomes such as time saved, decision quality and risk reduction. Explain how you would check whether observed use represents sustained value rather than simply a successful launch.

How would you incorporate security and governance requirements throughout AI product delivery?

What to coverSuggested preparation: Explain how guardrails and review needs would influence discovery, specifications, evaluation and rollout. Discuss trade-offs when a useful workflow cannot yet meet its acceptable risk or error tolerance.

How would you assess a workflow built with an approved low-code or agentic AI platform?

What to coverSuggested preparation: Use the posting's examples—Microsoft Power Platform, Copilot Studio or equivalent tooling—to discuss configuration, integrations, testing and deployment. Explain how you would judge whether non-technical users can operate the workflow reliably within approved boundaries.

When should a product-specific agent capability become a reusable platform component?

What to coverSuggested preparation: Describe how you would identify a pattern shared across products and communicate the requirement to platform engineering. Weigh reuse and consistency against the cost of generalising a capability too early.

How would you plan adoption after deploying an internal AI product?

What to coverSuggested preparation: Cover role-specific education, rollout, feedback loops and measures such as activation, usage depth and task completion. Explain how those signals would guide changes to the product or its enablement materials.

Behavioural practice

Tell me about a time you turned an ambiguous stakeholder request into a scoped product opportunity.

What to coverSuggested preparation: Use a real situation, the research and framing actions you took, and the resulting decision or outcome. Show how you handled competing interpretations without claiming certainty you did not have.

Describe a time you resolved a delivery blocker across technical and business teams.

What to coverSuggested preparation: Set out the situation, your actions to clarify dependencies and options, and the result. Focus on your contribution to a cross-functional decision rather than attributing the whole team's work to yourself.

Tell me about a time you communicated a system's limitations to non-technical stakeholders.

What to coverSuggested preparation: Give the situation, how you explained uncertainty or performance evidence, and what happened next. Choose an example that shows clear expectation-setting while preserving a useful path forward.

Describe a time a risk or governance concern changed a product plan.

What to coverSuggested preparation: Explain the situation, how you worked with the relevant partners, and the resulting change or decision. Show the trade-off between user value and acceptable risk without implying a particular Temasek approval process.

Tell me about a time you helped colleagues adopt a new tool or workflow.

What to coverSuggested preparation: Describe the audience and starting problem, the facilitation or enablement actions you took, and the observed result. Distinguish attendance or initial interest from evidence of continued use.

Sharpen your knowledge

Topics to prepare

End-to-end AI product ownership
The role owns discovery through deployment, adoption and improvement, with measurable outcomes. Suggested exercise: map a hypothetical research workflow from problem statement to sequenced roadmap, then distinguish launch metrics from sustained business impact.
AI design trade-offs
The posting calls for decisions on RAG, fine-tuning, prompt engineering, model selection and reliable agent workflows. Suggested exercise: compare those approaches for a hypothetical portfolio-monitoring task, stating the data needs, failure modes and reasons for each choice.
LLM evaluation and version monitoring
The role defines evaluation datasets and ongoing benchmarks, using automated evaluation, human review and red-teaming. Suggested exercise: draft a small test set and rubric, then specify how to compare model or prompt versions and detect quality degradation.
Business-to-AI specifications and governance
Specifications must cover data, integrations, error tolerance, prompts, agent behaviour, tools, API contracts, guardrails and feedback; governance is embedded throughout delivery. Suggested exercise: outline these sections for a hypothetical internal workflow, including where human review is needed.
Business value and adoption
The posting connects usage, accuracy and latency to time saved, decision quality and risk reduction, and calls for rollout and feedback loops. Suggested exercise: sketch a measurement plan separating activation and task completion from longer-term impact and error trends.
Citizen developer enablement
The role designs learning pathways, starter templates and agent blueprints for non-technical employees using approved platforms. Suggested exercise: outline a role-specific learning pathway and identify what a pre-validated starter template would need to make adaptation safer.

Skills to revise

Turn workflow research into a backlog
User research, workflow shadowing and prioritised backlogs are explicit responsibilities. Suggested practice: draft interview questions for an investment or operations user, then convert hypothetical pain points into scoped backlog items with prioritisation rationale.
Write an actionable AI product specification
The role translates business needs for engineers and domain specialists. Suggested practice: write one sample specification with data and integration needs, agent behaviour, acceptable uncertainty, guardrails, feedback capture and clear scope boundaries.
Run an evaluation comparison
Operating LLM evaluations is a stated requirement. Suggested practice: score two hypothetical prompt versions against the same cases using an explicit rubric; record human-review disagreements, red-team failures and a decision on whether to release.
Build and deploy an agent workflow
Hands-on workflow building and deployment with Microsoft Power Platform, Copilot Studio, Anthropic Claude or equivalent tooling is required. Suggested practice: if a suitable environment is available, build a small workflow using non-sensitive test data, add a human checkpoint, deploy it and inspect its outputs.
Facilitate a hands-on enablement session
The role runs workshops and office hours for non-technical employees. Suggested practice: rehearse a short session that explains a starter agent blueprint, has participants adapt it to a hypothetical task, and gathers feedback on usability and risks.

Your preparation checklist

Select a production AI product walkthrough
The role requires ownership from problem framing through deployment, adoption and improvement. Prepare one example showing your decisions, launch, measurable impact and subsequent changes; distinguish production use from a prototype.
Document your product-management evidence
The posting specifies 4–8 years in product management and at least 2 years owning AI or ML products. Prepare a dated timeline and concise evidence for those requirements, if applicable.
Practice a workflow discovery exercise
The role involves researching investment, portfolio monitoring, research and operations workflows. As preparation, sketch user-interview questions, pain points, opportunity sizing, a prioritised backlog and a first delivery milestone for one hypothetical workflow.
Work through an AI design decision
The posting calls for choices among RAG, fine-tuning and prompt engineering, plus reliable agent workflows and human checkpoints. Prepare a trade-off explanation for the hypothetical workflow, including failure modes and when a person should intervene.
Draft an AI product specification and evaluation plan
The role covers data and integration needs, prompts, agent behaviour, tools, API contracts, guardrails and error tolerances. Prepare a short sample specification with an evaluation dataset, automated and human review, red-team cases, and benchmarks across versions.
Show how you measure value and sustain adoption
The posting connects usage, accuracy and latency to time saved, decision quality and risk reduction. Prepare a sample metrics dashboard and explain how feedback, security and governance, rollout, and cross-functional dependencies would inform improvements.
Prepare a citizen-developer enablement example
The role includes hands-on platform workflows, starter templates and workshops for non-technical employees. If relevant, bring a walkthrough using Power Platform, Copilot Studio, Claude or equivalent tooling, plus a short workshop outline and an adaptable agent blueprint.
Write questions for the interviewers
Ask which internal workflows are the immediate priorities, how product evaluation and governance are handled, what platform capabilities are reusable, and how adoption and business outcomes are measured. These are questions to clarify, not assumptions about Temasek’s processes.

Availability last checked: 4 Nov 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 Singapore. Matched to this role, not your profile.

Explore more jobs

Finding similar opportunities…