The Creation of Lisa — How an AI Coworker Was Designed
Part 1 · Originally published January 2025
The story of Lisa starts with a central idea: an AI coworker needs a dedicated intelligence layer — what we call Recall Intelligence — to operate quickly, reliably, and intelligently in real-world work environments. But what exactly is an AI coworker — and why is it so promising for supporting people and organizations?
1. Multimodal communication: interacting naturally and effortlessly
A core requirement for Recall Intelligence is that the AI isn't limited to a single form of interaction — not just chat, not just voice. People communicate across many channels: email, messaging, video calls, project management tools. An AI agent like Lisa needs to:
- handle speech, emails, and chats equally well,
- integrate with multiple platforms and formats,
- adapt to changing contexts (smartphone, web interface, ERP system).
Why this matters
The more naturally the AI interacts, the faster users accept it as a colleague on equal footing. Only an AI that works smoothly inside existing channels can truly reduce daily workload and accelerate business processes.
2. Executing complex workflows alongside humans
An AI coworker isn't about automating individual tasks; it's about understanding and orchestrating entire workflows. That matters especially in supply chain management, where multiple interdependent processes have to line up.
Key capabilities include:
- coordinating sub-tasks (demand planning, supplier management, inventory control),
- transparent communication — the AI explains why it recommends a given decision,
- role awareness — the AI acts as a team member, complementing human expertise rather than replacing it.
Why this matters
When an AI understands the bigger picture, it can spot issues early, make informed suggestions, and provide real decision support. That moves it beyond a simple automation tool toward a trusted collaborator.
3. Learning from humans — and keeping what it learns
One of the defining questions in Lisa's development was how the AI should learn. Traditional machine-learning models are trained offline and stay relatively static until the next update. An AI coworker works differently:
- it learns from every interaction — each email, chat, and piece of feedback refines how it works,
- it adapts to context — company-specific processes, conventions, and best practices,
- it enables a two-way exchange — people give feedback, the AI absorbs it, and its working knowledge grows.
Why this matters
This is where the real compounding happens. The genuinely valuable work in supply chain isn't the routine case — it's the exception: the confirmation that disagrees with the order, the gap that needs a substitute, the situation no rule anticipated. When a human resolves one of those, Lisa retains how it was resolved, so the same exception doesn't have to be solved from scratch twice. Over time the pile of cases that need a human shrinks instead of resetting. A self-learning coworker that captures resolutions, not just data, embeds itself into a team and gets structurally better the longer it runs.
4. Access to real-time knowledge and decision relevance
To excel, Lisa needs a steady feed of current data — inventory levels, supplier updates, customer signals — and the judgment to know what matters. The AI has to process information quickly, prioritize, and factor it into its decisions.
Key features include:
- relevance-focused decisions — the AI discerns what needs immediate attention,
- condensed information — prioritizing the essentials so they fit within its working context,
- automated logging — each decision and update is recorded, enriching the system over time.
Why this matters
Any modern AI is only as good as the data it takes in. By filtering and contextualizing effectively, the AI gives targeted, real-time recommendations that fit the workflow at hand.
Inspired by how the brain works
The design of Recall Intelligence also borrows, loosely, from how the human brain operates. We don't claim to replicate neurophysiology — but a few principles translate well as design metaphors:
- Parallel processing — handling multiple data streams at once rather than one at a time,
- Filtering signal from noise — tuning out the irrelevant to prioritize what's critical,
- Working context — applying situational knowledge so responses are accurate and situation-aware.
Why this matters
Systems built around these principles prioritize more effectively and adapt better to fast-changing environments.
Why Lisa became an AI coworker for purchasing and planning
Once the foundational architecture for AI coworkers was in place, our roots in supply chain management took center stage. The capabilities we'd been building — multimodal communication, workflow orchestration, continuous learning, real-time awareness — turned out to be an ideal match for the complexity of the field. Why?
- Many interaction points — suppliers, logistics providers, and internal departments all have to stay in sync,
- Strong automation potential — repetitive work (orders, confirmations, follow-ups, planning) can be handled by an AI coworker,
- High impact — even small gains in forecasting or exception handling translate into meaningful savings in cost, time, and service level.
So Lisa took shape as an AI coworker for purchasing and planning — anticipating demand, checking order confirmations against the purchase order, detecting and resolving deviations, watching supplier performance, and proactively managing order proposals. Over time, she keeps learning how to handle these processes better.
Conclusion: a digital colleague, designed to compound
Giving an AI a name and a role is more than a marketing choice. It's a way to build an agent that operates holistically rather than as a collection of disconnected functions.
By building Recall Intelligence on multimodal communication, workflow orchestration, continuous learning, and context-aware decision-making, an AI coworker grows into something closer to a digital colleague — one that complements human expertise instead of replacing it.
That foundation is what lets an agent like Lisa excel in supply chain management today, and what will let the same approach extend to other fields tomorrow. It enables authentic interactions, which in turn drive more effective collaboration and lasting value for organizations.
