Specialist technology firm, Lightworks, along with Scotiabank, Sun Life and TELUS have banded together to create a new artificial intelligence (AI) consortium to build and govern the infrastructure needed to implement AI, agentic AI in particular, in a safe and responsible way at scale.
A July 7 announcement notes that many of the core challenges the consortium hopes to address are shared, including the integration of diverse standards, platforms and technologies, along with governance, operational control and oversight requirements which emerge as AI use scales. The group intends to pool its engineering resources to conduct research and jointly build “mission critical” AI control systems and intellectual property (IP) which would otherwise need to be developed independently.
The resulting IP will be deployed individually. Members will retain perpetual use and ownership rights.
“They own the IP and can have a perpetual, ongoing right to use that IP for their own organization,” says Sun Life’s executive vice president, chief information and technology innovation officer, Laura Money. “We’re all building this together. We all get to use it forever without having to pay licensing fees to a third party.”
Flagship program
The consortium’s flagship program, the Agentic Control Plane or ACP, is already running in production at TELUS, giving the company the visibility and control needed to manage AI agents. (Those which operate and execute activities autonomously.) “It helps support regulator compliance, maintain operational control and currently processes more than two trillion tokens per month across member organizations,” they write.
At Sun Life, Money says the company plans to have three versions of the ACP, one for its U.S. business, one for Canada and one for Asia, with local customizations made for each jurisdiction.
“The opportunity of the AI consortium is to help us move from experimentation to responsible, production-scale impact,” she stated in the consortium’s announcement about its formation.
In a discussion with the Insurance Portal about the consortium’s development – each member is dedicating five full-time engineers to the effort and has contributed an undisclosed sum each to fund Lightworks’ engineering efforts – Money says the ability to control agents is something that no vendor today has created a solution to address. The ACP is designed to track what the agents are doing along with the systems and data they’re accessing. It is also designed to manage the agents such that their activities are halted for human inspection if they appear to be doing things they shouldn’t.
“The flagship of the consortium is the ACP. Every organization is going to need that, particularly every regulated organization that has any kind of sensitive data – they’re going to need that. Building it together is way better than each of us trying to build it on our own,” Money says.
Canada is lagging in AI adoption
She adds that Canada is lagging AI adoption compared to some other countries. “We see this as an opportunity to really help some of those industries,” she adds, noting that the consortium hopes to add members going forward, “but it’s not fatal if it doesn’t happen either. We’re quite happy working with the four of us, as well.”
Going forward, she says the consortium’s developments could be limited by the imagination only. “It could go in all kinds of directions,” she says. Could the industry use its own custom large language model (LLM) encompassing banking, telecom and insurance regulation in Canada? It’s not something talked about, but as a collective absorbing the development, such possibilities become more reasonable to consider.
“I think we have to take it step by step,” Money says. “Let’s get our control planes up and working. As a construct, if it works, I could see it going in many directions.”
Among the things that excites Money about the project is being able to ascertain the health of the agents being deployed. “Because it’s one thing to control them, it’s another to say are they really creating the value they should be creating? Are they really doing the things that they should be doing at the frequency we thought they could?”
Customer-facing activities
Currently she says AI agents are in use currently in the coding space. “That’s great. We need a control plane for that. But as we start using agents to support more customer-facing activities, I think it’s going to be very important that we understand exactly what they’re doing and how they’re doing it.”
The development of the ACP is expected to help the regulatory effort, as well. “This only helps to answer all of the questions that regulators will inevitably have,” she says. “They’re looking for evidence that you’re doing this safely. How do you know the models are doing the right things?”
In short, the company emphasizes that its efforts are ongoing to bring about AI development in a responsible and safe way and that working together with those facing similar challenges is one way to get there. “There are things we can do together without giving up competitive advantage,” she says.
