AI, Large Action Models and the future of the digital workplace
For decades, the digital workplace (DWP) has been seen as a tool for doing business. Innovations in unified collaboration, communication, service desks and other areas have been significant, but they have merely been business enablers. This is all about to change, as new developments like Large Action Models (LAMs) have the potential to put AI innovations at the core of the business itself.
What is a Large Action Model (LAM)?
Until now, Gen AI has acted as a sort of intelligent assistant, or at best, a teacher — performing routine tasks like note taking, drafting emails or summarizing documentation. Despite its ability to free-up time for more strategic or creative work, it has always stopped just short of the actual doing.
That’s where LAMs come in. A Large Action Model is not just an assistant, but an actual agent capable of executing tasks. Just as Large Language Models (LLMs) can sort through vast amounts of textual information to determine meaning and intent, LAMs are designed to analyze, understand and replicate actions. Instead of just supplying a list of steps required to accomplish a task, it can actually interface with other digital or physical systems to follow a task through to completion based on the user’s instructions.
Gen AI is a catalyst that is transforming business data into a democratized, accessible fountain of knowledge. However, the true transformation is in translating these insights into action with Large Action Models.
Potential use cases for LAMs
One example we are likely to see in the near future is a virtual travel agent. Imagine a scenario where you provide a Gen AI agent (an LLM) with a list of preferences and parameters about what type of destination you would like to visit and when. It returns a shortlist of possible options, and you interact with it to understand the pros and cons of each — including the accommodations, amenities and activities. You might even use a VR headset to take a virtual tour. If you’re a Brit, naturally you opt for the all-inclusive resort, because who wants to make that many decisions?
Once you have selected your destination, the details are handed over to a LAM-based agent. Over and above finding the most affordable flights, it can actually book the tickets and make reservations on your behalf, using your stored credit card data and profile information. While making the arrangements, it takes into account your budget, schedule, room preferences, dietary needs and other parameters.
Another potential use case is in procurement, where a LAM could develop a tender based on your business requirements, gather a list of suppliers, float an RFP and evaluate bids. It could also potentially automate the process of reviewing supplier contracts, negotiating prices and submitting purchase orders.
Farther down the road, Large Action Models could be entrusted with physical tasks, such as controlling factory operations to automatically shift the product mix to meet real-time demand. In healthcare, skilled clinicians could be better leveraged by letting a LAM automate important but mundane tasks such as preparing documentation, patient monitoring and administering care plans.
The only limitation is how well the models are trained, what safeguards are in place, and the level of trust placed in the LAM.
The business impact of Large Action Models
As the concept of Large Action Models evolves, the impact will be felt across every industry. Intent detected by LLMs will be used to autonomously execute complex sequences of actions across distributed digital platforms, applications and even the physical world. The result is that GenAI will evolve from a teacher to a doer.
Right now, AI agents can “join” your meetings, transcribe the conversation, take notes, and summarize the outcome and action points. As convenient as that is, all AI can do right now is provide you with a list of tasks — the “doing” is still up to we humans.
Using LAM, however, that same agent could schedule and book a room for a follow-up session, email pertinent questions to individuals not in attendance, or even order lunch for the entire group if the meeting runs over. This will rapidly evolve into fully fledged multi-modal digital colleagues participating in meetings, digesting, interacting and contributing in real time — without worrying about whether they are checking their Instagram accounts during the meeting.
We could even imagine multiple digital colleagues from different vendors participating in the same meeting, so at best it increases confidence and provides a different perspective. At worst, it will be a fun experiment.
Where will AI take you next?
Gen AI is not just an addition to the digital workplace; it is a catalyst that is transforming immense amounts of existing business data into a fully democratized and accessible fountain of knowledge. For more reading on how Gen AI is being used to spark innovative thinking, I highly recommend this insightful article from Harvard Business Review.
Presently, the availability and quality of data are critical success factors for Gen AI, but the true transformation will take place when we can translate these insights into action. With Large Action Models, digital workplace technologies will finally transcend their status as an enabler to truly become a core component of the business.
Given the immense future value at stake, the degree of integration required to incorporate Gen AI, LLMs, LAMs and core business applications should increasingly drive the selection of future DWP partners.
My colleagues at Atos have been doing a lot of thinking about how AI is reshaping the digital workplace, improving enterprise knowledge, empowering software coding, and paving the way for autonomous business.
If you are considering how to leverage the latest AI innovations, or how to proceed with the next phase of your digital workplace transformation, we are here to listen, to guide you, and to help make it a reality.
Posted on: July 24, 2024
Adil Tahiri
Head of CTO and Client Advisory GroupMember, Atos Research Community
View detailsof Adil Tahiri>
