Introducing Memory in Sofie

Memory brings your preferred working style, institutional context, and best practices into every AI output. The longer you work with Sofie, the better it knows how you work.

TransferAI Team

Biopharma operations are built on continuity. What was validated before shapes what gets executed next. The batch record references the process development report. The tech transfer protocol builds on the sender site's history. The deviation investigation draws from every prior corrective action.

In a line of work built upon institutional knowledge, this is all but surprising. Biopharma teams work to become experts and understand the full breadth and depth of the site, the product, the regulatory landscape and the way their team prefers to work. With those expectations for their team members, our users are coming to expect that same vigor from Sofie. That's why we built memory.

Explaining Memory

Memory is a persistent, intelligent layer that carries forward the context you choose including your role, your products, your processes, your working preferences, and the regulatory frameworks that govern your work. Memory lays a foundation so that every conversation with Sofie builds on the ones before it.

There are five components:

  1. Memories capture facts, context, and preferences over time. As you work with Sofie it learns the details that shape how you work from your team structure and product modalities to site-specific SOPs and formats that work for your organization. These aren't stored passively — Sofie uses a hybrid retrieval system that combines semantic search, entity matching, importance weighting, and temporal awareness to surface the right context at the right moment.

  2. Directives are persistent rules that always apply. "Format all outputs for GMP documentation." "Always include regulatory references in compliance deliverables." "Use formal language in quality reports." Set them once and Sofie will follow them in every conversation, without being reminded.

  3. Episodes are episodic memories of your work sessions over time. Sofie remembers not just facts, but context — the tech transfer review from last Thursday, the deviation investigation you walked through together, the regulatory submission you drafted in three sessions. Episodes give Sofie continuity across conversations so you never have to re-explain where you left off.

  4. Knowledge Graph is a structured map of how your world connects. Sofie links entities so context isn't just stored, it's understood relationally.

  5. Analytics gives you visibility into how your institutional knowledge is being used. See which memories surface most often, which directives are applied, and how Sofie's context awareness evolves with your team.

Why Memory in Sofie Is Important

Other AI platforms treat every user the same way no matter your role or motive. But biopharma work is not uniform, and neither are the people doing it. That's why we built memory to close this gap. When applied correctly, it means outputs that already reflect your regulatory framework, your documentation conventions, and the products you support — without requiring a preamble every time.

Here's what that looks like in practice:

  • A QA lead reviewing environmental monitoring excursions gets outputs grounded in your site's deviation history, monitoring thresholds, and quality documentation conventions

  • A process engineer running scale-up analysis gets responses informed by your existing process parameters, prior tech transfer decisions, and manufacturing site context

  • A regulatory affairs specialist drafting a CMC submission gets outputs that already reflect your regulatory framework, filing conventions, and product-specific requirements

  • A validation lead building cleaning protocols gets continuity across prior validation work, equipment configurations, and your organization's documentation standards

Every interaction adds context that makes the next one faster, more precise, and more aligned to how your team actually works. In regulated environments, where consistency is not a preference but a requirement, this kind of continuity is what separates a useful tool from an indispensable one.

Memory does not cross organizational boundaries, retain information without explicit user or admin authorization, or persist beyond your organization's defined retention window.

Governance as a Central Premise

Governance is a foundational principle of memory, not a secondary consideration. Organization admins have full authority to enable or disable memory at the user, team, or site level, ensuring complete control over what the platform learns and retains over time. In practice, when memory is active, Sofie can reference past workflows, prior decisions, and accumulated institutional context within your organization's defined retention window to better inform a new task. Because Sofie operates inside GxP-regulated environments where traceability is a baseline requirement, every memory applied within a workflow is logged, attributable, and fully auditable. We will continue partnering with quality, IT, and compliance stakeholders to deepen granular administrative controls, user-level transparency, and data segregation capabilities as the Memory feature continues to mature.

Three Ways to Get Started with Memory

Memory is available now. To enable it for your organization or explore configuration options, visit your workspace settings or reach out to your TransferAI account team. Once it's on, building your memory is as effortless or as deliberate as you want it to be.

There are three ways to shape your memory:

  1. Sofie learns automatically. As you work, Sofie picks up on the context, preferences, and patterns that shape how you operate and captures them in the background without any manual input required.

  2. Add them manually in settings. Go to Memory under Settings to add, edit, or remove specific facts, preferences, or context.

  3. Instruct Sofie directly in chat. In any conversation, you can tell Sofie to log something: "Remember that we use a two-stage review process for all deviation reports" or "Save my preference for executive summaries with risk matrices." Sofie will capture it and carry it forward.

If you want a fresh set of high-impact ideas tailored to your use of Sofie, request a 1:1.

A pilot that works on day one

A pilot that works on day one

In one hour, your team is live. One week to see your work multiply. Sofie is built for biopharma teams from day one, not day ninety. Your first result comes from your first conversation.

In one hour, your team is live.
One week to see your work multiply.
Sofie is built for biopharma teams from day one, not day ninety. Your first result comes from your first conversation.

In one hour, your team is live. One week to see your work multiply. Sofie is built for biopharma teams from day one, not day ninety. Your first result comes from your first conversation.

Don’t change how you work,
change what you’re working with.