Takeda mAI Access | Nimble Showcase

One intelligent workspace between evidence and reimbursement

Takeda’s medicines only reach patients once regulators accept the evidence and payers accept the price. That second step, market access, is getting harder: HTA bodies scrutinise data more closely, and every country runs its own dossier logic.

The work was spread across SharePoints, vendor reports and a dozen disconnected AI tools. Takeda’s market access teams needed one place to work, and a way to turn a growing estate of AI agents into something analysts would actually use.

As Takeda’s strategy, design and engineering partner on mAI Access, we shaped the vision for an AI-native one-stop shop, prototyped it to CEO level, and engineered the experience layer that orchestrates Takeda’s AI capabilities: the One-Stop Shop, the embedded Asset Wizard, and the backend-for-frontend architecture that ties them together.

opportunity

For market access analysts

Finding existing evidence meant an email-driven hunt across SharePoints

Answering one payer question could mean reading a 6,000-page clinical study report by hand

Global templates rarely matched local HTA requirements, so affiliates rebuilt dossiers from scratch

Scenario building meant weeks of rework and slide shuffling before strategy could even start

Useful AI tools existed, but they were hidden, fragmented and inconsistent, so few people used them

For Takeda (Global, Regions & affiliates)

HTA bodies and payers grow more demanding of the data with every submission

Every major pharma company invests in GenAI; Takeda needed adoption beyond the pilot stage

Several squads and vendors built capabilities in parallel, with no shared experience layer holding them together

The ambition: a single orchestrator for all mAI agents, serving ~400 market access users

Continuous guidance

One journey from evidence to reimbursement: mAI Access maps every tool onto the stage where it helps

solution

Five tracks from vision to a live, AI-native platform

Discovery: framing the one-stop shop

Ten stakeholder interviews mapped every AI tool in Takeda’s market access landscape

Each tool charted with its pain points, owners and future role, from Evidence Gateway to SLR automation

Defined the north star: an agentic, intent-driven front end that orchestrates capabilities instead of listing them

Adoption first: findability in daily workflows, local usability and the terminology analysts actually use

The CEO prototype

A clickable, end-to-end prototype of the ideal state

A personalised home that knows each analyst’s assets and priorities

An always-on assistant next to a full Insights Generator flow

From HTA summary to gap analysis to scenario planning in one journey

Secured CEO buy-in and additional budget for the programme

Global Value Dossier builder

Template-led assembly with claim selection, evidence anchoring and review built in

Endorsed global content kept separate from localisable sections

Versioning, audit trails and export to payer templates such as AMNOG and NICE

Presented in Zurich and adopted as the first-module definition

Shipping the One-Stop Shop

Rebuilt a static, SharePoint-like tool library into a modern React workspace

Role-aware navigation, tool categorisation and bookmarks for novices and experts alike

A market access timeline that maps every tool onto the stage where it helps

Extended Takeda’s Tailwind design library with the conversational components it lacked

Delivered in weekly sprints with weekly demos, perf budgets, security guardrails and telemetry

The Asset Wizard, embedded

Conversational Q&A over PubMed, EMA and internal sources, native inside the platform

Every answer arrives structured:

Sources and confidence scores

Backend-for-frontend layer

We shipped the first BFF building blocks: a secure document proxy and typed API contracts

And architected the full BFF, assistant orchestration and integration layer

One contract for the UI instead of a patchwork of team-owned endpoints

Capability teams plug in behind clear contracts, users see one product

Backend-for-frontend architecture

In AI products, the user experience is shaped as much by backend behaviour as by the interface. A slow agent loop, an inconsistent response format or conversation state in the wrong place breaks the experience, however good the screens look. So we treated the layer between the UI and Takeda’s AI capabilities as product work.

A React and TypeScript workspace talks to a UX-driven BFF in Node.js that shapes, aggregates and secures calls to multiple capability backends. An assistant orchestration layer holds intent and session context, routes tool calls and enforces guardrails on Takeda’s approved LangChain and LangGraph stack. Contract-first plug-ins connect Evidence Gateway, the Asset Wizard and Insights services on Takeda’s AWS and Databricks foundation.

The principle: capability teams build what the system can do; the experience layer decides how users experience it. Design sprints produce intent maps and API contracts first, so a multi-vendor programme still ships one product.

Outcomes

Delivering real value

A shorter road to reimbursement

Every week between approval and a reimbursed price is a week patients wait. Evidence that took an email hunt is one question away, and HTA landscapes, gap analyses and upside scenarios that took weeks of manual synthesis are generated in hours. Time-to-first-draft dossier drops, and Global and local teams align on strategy earlier in the launch.

Less reliance on external vendors

Literature reviews, HTA environment scans and dossier drafting were long outsourced: costly, slow, methodologically inconsistent and often outdated on arrival. Moving that work into in-house AI capabilities behind one experience layer lets Takeda keep the knowledge, control the method and cut recurring vendor spend, while its own analysts stay in charge of the judgement calls.

“[Client quote to be sourced and approved by Takeda]”

[Name], [Title] Takeda