Using AI in Emergency Management: It's Time

?

Innovation. That seems to be a prevailing theme in my recent string of articles this last while, and why wouldn’t it be? I like it. I appear to practice it often. And I am a firm believer that innovation is foundational to any organization, private or otherwise, adapting to the incredible rates of change we’re experiencing. Much of that change is a result of a near perfect storm of emerging and converging technologies, mixed with a good healthy dose of really, really smart people. So, here’s an example of an innovation that has occurred when I happened to engage one of those really, really smart people. And believe me, there’s more innovations to come. Because I also found another source of really, really smart people. More on that in a moment but first, let’s explore what happens when an existing innovative practices gets some AI thrown at it.

To start, let’s establish a basic ground rule around the use of AI-enabling technology. It isn’t unique to me. In fact, it is fundamental to anything to do with AI. At the core of it is the notion that AI can and should be used to enhance what is typically some sort of manual, or labour intensive and/or time consuming practice. Fun fact:? if you google anything to do with stats around the use of AI for increasing productivity in the workplace, you will find no shortage of published studies that generally indicate that improvements in output and throughput are quite high, usually somewhere north of 60%. That’s an important fact to remember. It seems I’m not the only one that realizes the benefits of AI for enhancing productivity in the workplace.

Now, this isn’t about AI and robots replacing all the workers. It’s about the organizations having to adapt their work practices to keep up with the increasing demands (more climate change related disasters for example). When I was first looking into AI a couple of years back, I reached out to a number of clients to get their thoughts. One bit of feedback from a friend who manages a regional emergency program really stood out for me, and my approach to AI enablement is very much focused on it.

A prevailing theme in a lot of local government and First Nation emergency operations centres in recent years has been burnout and increased stress. I would offer that the same exists in any level in any organization’s emergency operations. In fact, we’ve seen lots of PTSD starting to appear. It has always been there, but now there are more emergency incidents and events requiring more personnel to manage over longer periods of time. Emergency management workers are definitely starting to experience a cumulative fatigue and stress. So it was in this context that my friend offered this comment:

“if you can develop anything that allows me to use less people in my EOC for shorter periods of time, I am all in”.

He, like so many others with their EOCs, has been experiencing a staffing capacity challenge related to stress and burnout. The Coles Notes on that is so many people are starting to burn out because of the additional stresses put on them when they leave their regular job to do a stint in the EOC. For some people, there comes a time when it just isn’t fun anymore. For the record, I still enjoy it. But that landed with me because I’ve been one of those people in the EOC for very long days over extended periods of time. I realized then that AI can be a good thing, and even a necessary thing.? So, bringing this back to the really, really smart person that I met, I presented? him with the challenge of figuring out what AI could do to enable some of the business practices around a methodology called PPOST.

I’ve written on PPOST before, and will continue to do so, because I have found it to be a very powerful planning tool for a number of things, not the least of which is things like Advance Planning and initial Recovery Planning. To me, PPOST is already an innovation because it builds on the management by objectives that is taught in all levels of the Incident Command System (ICS). To remind you, PPOST stands for:

• Priorities – set your community priorities
• Problems – identify the issues and challenges specific to your community interests
• Objectives – what needs to be done; set them based on your priorities & problems
• Strategies – how you are going to do it; set them to achieve your objectives
• Tactics – who is going to do what by when; set them to achieve your strategies

We’ve added a couple of our own innovations to the PPOST methodology whereby we can use it to develop an initial recovery plan for fairly large and complex events in less than 72 hours. But in the context of this article, what we thought we’d take a look at is how AI might enable the process. Remember, the fundamental consideration around using AI is that it must be used to increase productivity, which ideally reduces staff time commitments. So, in this case we think that we’ve landed on something that will do just that.

Let’s use the development of an initial Recovery Plan as an example. We’ve done several large wildfire and flood events with pretty good success. To date, the process has looked something like this:

• Event happens
• Small team assembled to facilitate PPOST session & write report (2 -3 people)
• Facilitated inquiry session using PPOST to identify all the considerations around what the initial concerns and challenges are
• Draft report written and shared with the client; this includes objectives, how they are going to be met, who is going to do what by when, an initial org chart, a list of resources required and their source, and an estimated budget to get the recovery operations set up
• Final report submitted to higher level government for funding approvals to establish and operate a recovery management team and operations centre
• Recovery Manager and team start on initial immediate and short term objectives; start engaging on medium and longer term planning

As I mentioned, we’ve landed on a way to turn out that initial Recovery Plan in less than 72 hours. That’s because we’ve streamlined the PPOST inquiry so that it is facilitated by someone who understands all the moving parts to disaster recovery, and we’ve developed templates for each disaster type. That’s a common sensical approach I would think. But we’ve also added a couple of pieces to that. As we are facilitating the inquiry, we are also creating an recovery operations centre org chart, and identifying all the resources and how they are going to be procured. Because we’ve sorted out the “who is doing what by when” piece during the facilitated inquiry, it’s a pretty easy step to create an initial budget estimate. And we’ve rolled all of this process up into a checklist based step-by-step playbook.?

All of this done in less than 72 hours, which we’ve come to realize is typically much quicker than the classic approach to recovery planning. Remember, this is about developing the initial recovery organization and getting it going in a few days. Once the first few members of the team are in place, the PPOST methodology can be used to identify all future iterations of the planning. We think that the 72 hour thing, which can often be less than 48 hours, is an impressive amount of time to turn out an initial recovery plan. So you might be asking yourself “why use AI to speed up a process that is already pretty quick?” Good question. Here’s my “why”.

What we’ve discovered over the last 3 or 4? years that we have been using PPOST for initial recovery planning, and other applications as well, is that the real strength in this methodology is in the facilitation. More specifically, in the facilitator. The PPOST approach the way we’ve laid it out is very methodical and linear in its flow. Pretty much anyone can follow it and have some success, especially if they use the 5Ws (who, what, where, when, why and sometimes how) in their inquiry. But not everyone has the ideal blended background to really dig into it and know what’s hidden around every corner or buried under every rock, euphemistally speaking. That ability to tease out the nuggets of wisdom. What we’ve seen works very well is when you have a facilitator who has a lot of experience around site command, EOC leadership (community, regional, provincial/state, including First Nations), the practices of assisting and supporting agencies, and a thorough working knowledge of all relevant legislation, policies, and funding programs. There’s not a lot of those folks around, and we know that. Hence the AI-enabled approach to PPOST.?

Because not everyone has that ideal blend of skills to do a really thorough facilitation, but most have exceptional skills sets to bring into the process, we figured AI might be able to help them a bit. So, we built an AI tool around:

• A large language model (LLM) on a dedicated and secure server, located here in Canada for legislated reasons
• A series of prompts based on the inquiry methodology
• A knowledge base comprising data from previous actual PPOST sessions

The idea is that as someone is facilitating the inquiry, their support person is capturing all of the information into the system. While that is happening the system is generating suggestions for the facilitator to address with the small EOC planning group that has been assembled. It can guide the inquiry with enough accuracy to start with so that the system-generated responses can either be accepted as they are, and/or modified, and/or completely new content can be added as it is identified through the facilitation.

Right about now you’re probably thinking “cool, I’d like to take a look at that”, and you will be able to one day soon, but what you need to know is that this is still version 1.0 alpha testing stage. We have already started running it past some of our emergency management peers to get ?their take. Not surprisingly, the few that have already started to look at it have provided us with a bunch of “can it do this” or “what if it could do that” kind of feedback. Classic. But we already have a goal for version 2.0 where ideally, the facilitated PPOST session is recorded, transcribed, fed into the system, and a draft initial recovery plan is generated before we even leave the room.

Ambitious? Yes. Innovative? We like to think so. Possible? Yes. All the bits and pieces are there in AI world to make this happen, and we’re working on that. Is this going to be a game changer or disruptor in the emergency management planning world? Who knows. Will it save staff time and potentially cause less stress on some people? We sure hope so. If there’s one thing we’ve come to realize over time, it’s that the sooner you can get the supports to people impacted by an emergency event or natural disaster the quicker they can get to their “new normal”. We’ve also come to realize about most people who work in any form of public service, particularly emergency management, is that they are there to serve the public. Many of them tend to wear the event on their sleeve, so to speak. They buy into their role with great emotional attachment. So, anything we can do to help them help others can’t be a bad thing. And we’ve had a few more “aha” moments since we began this project.

It seems like there are a number of business practices within the emergency management domain that can probably benefit from AI enablement. We’ve got no less than 7 identified that can form part of an integrated suite of tools. The really, really smart guy that we engaged to build out the AI-enabled PPOST tool has taken an initial look at each use case and said “doable”. What exactly that looks like, I can’t say for sure yet. Ask me in 6 months. But I can tell you that we’ve already engaged some really, really smart university students on a project to use AI to develop exercises. They’re doing it as part of their course requirements and so far, all it has cost is the price of taking the team out to lunch every couple of weeks. University students appear to like it when someone buys them lunch. With this project, in version 1.0 we are focusing on tabletop exercises that are based on existing master series of events lists and scripted injects from who we consider to be the most preeminent exercise and delivery company in Western Canada. What’s been rewarding for me is watching our really, really smart AI guy mentor the really, really smart computer science students through the process. I think they’re getting more than just a free lunch out of the deal. Anyway, version 2.0 will have a whole bunch more of AI-enabled processes and not surprisingly, our AI guy has said “doable” to this as well.

So, to wrap this up my guidance to you would be to start thinking about how you can improve on some of your more manual business processes. AI and automation software can certainly play a role in that but remember, PPOST is simply management by objectives on steroids. And the practices that we added to PPOST originally had nothing to do with fancy new technology. That said, the future is in AI-enabled practices so if you want to get a bit of insight here’s a link to a webinar that I recently attended called Practical Applications of AI for Emergency Management. It was hosted by a group on the West Coast of the USA who organize an annual conference called Partners in Emergency Preparedness (PIEP). Again, here’s some more really, really smart people doing what they love to do. Go check out the webinar recording.

?

?

?

?