A Strategist's Guide to GenAI Integration: the business of problem solving

Welcome back to my series of pieces all about generative AI and the opportunities we have to integrate them thoughtfully and impactfully into our teams and organizations. If you want to catch up on the first entry (and I definitely think you should!) you can find it here.

Let's start part 2 by talking about my whimsical fave: the Horniman walrus.

In 2016, Jim Carroll wrote a delightful piece for The Guardian about communication platforms using the Horniman walrus to illustrate his point. For those whose lives have not been enriched by the Horniman Museum, it's found in south-east London and is filled to the brim with art and artefacts and musical instruments acquired by Victorian tea trader Fredrick Horniman. It also houses an impressive natural history collection, part of which is a very unique walrus.

It turns out that when the walrus arrived in 1880's London, the taxidermists had never actually SEEN a walrus. But they had a job to do; the fact that they weren't quite sure what the end product should look like (and that it would be pretty hard to make changes once they were done) was less relevant than their looming deadline.

Dearest gentle reader, they overstuffed the walrus.

Mr. Carroll uses the Horniman walrus to make an excellent point about communication platforms that still holds true today:

How many of us operating in marketing and communications over the years, working with new technologies and media, have created our own Horniman walruses? How often have we endeavoured to put a new platform at the heart of our plans without being entirely confident what best to do with it?

My favorite thing about this piece is the metaphor, obviously. But a very close second is the fact that you could add AI into this exact conversation without missing a single beat. If there's one single thing any of us knows for sure, it's that gen AI has a lot of energy and momentum behind it. If your LinkedIn feed is anything like mine, there are videos every day promising to teach you "the number one hack to get the most out of GPT" and slide reels counting down the "tips and tricks to make Copilot the assistant of your dreams." There are definitely some exciting opportunities for brands to make use of this technology, but your Friendly Neighborhood Strategist is here to remind you that integrating AI into your stack and workflows without having a clear problem-solving objective in mind will. Cause. Problems.??

Here's my best advice: Be in the problem-solving business, not necessarily the AI business.??

Do you have an AI problem, or an automation problem? Are you using AI because it's the best tool for the job, or because the FOMO is real? There are opportunities to use AI to great advantage, but you shouldn't be using it just to say you are. Find a specific place AI can make an impact in your organization.?

Once you’ve targeted the specific problem, format your approach as you would any other major business objective. The best business objectives are SMART, and integrating generative AI into your processes should take the same approach.?

Specific – decide where generative AI can make the most impact and be as specific as possible when outlining those use cases. Chatbots, fraud detection, and content generation are some of the most common areas, but there might be goals that are more specific to your business such as product design or SEO efforts.??

Measurable – define what success will look like and make sure that you have the metrics and data to gauge and report on your success. Productivity gains, cost savings, or waste reduction are common examples, but your metrics may be more granular and/or specific to your department, team, or company. It’s important to keep in mind that your metrics might be tied more tightly to internal processes than external results – generative AI can be very helpful with increasing personalized experiences and powering recommendations based on A/B testing, and success metrics for those use cases might include increased click-through rate (CTR), decreased bounce rate, or improved scroll depth. Make sure you’ve established relevant KPIs for each use case you implement. Success speaks more loudly than words – make sure you’ve got the numbers to prove that what you’re doing is working.??

Agreed – business objectives should, ideally, impact more than one corner of the company. All stakeholders should agree on which business objectives generative AI will be best placed to impact to avoid conflict and, in the worst case, failure.?Consult with multiple stakeholders across the business to gather input and build support for the organized adoption of generative AI. Discuss what departments are likely to face the largest changes and which teams will have more time to access and understand the tooling available. Bringing as many team leaders as possible into this process is the best way to gain support and spot possible problems before they become large issues. Seek out diverse voices within your organization.??

Realistic – make sure that your goals are grounded and possible. There’s so much attention being paid to generative AI now and the fear of missing out is very real. That can easily impact business decisions. Be bold in your dreams but realistic in your short-term expectations to set your team up for success.??

Timebound – be sure to set a time limit and understand what success will look like for your business within that time limit. This is another time to be realistic - Steve Jobs pointed out that “We always overestimate the change that will occur in the short term and underestimate the change that will occur in the long term. People overestimate what can be done in one year, and underestimate what can be done in ten.”??

Choose your fighter wisely

Generative AI is a tool like any other; it’s not a silver bullet or a magic wand. Like personalization (and to a much lesser extent, the metaverse) before it, what you get out of it will depend on what you use it for, how you measure success, and how you learn from failure. One of the better times to answer those questions is when you're selecting the generative AI tool you're going to use.

This is a great time to listen to your team members - find out what they're using, especially if they're using something "off the books" that they find particularly valuable. Consult your IT team, who is going to have fast and accurate answers about whether the tool your considering will play nicely with the rest of your stack. Do a bit of market research and ask your friends and contacts if they've ever had trouble getting Tool Z through procurement (spoiler alert: you're going to hear 'oh yeah' a lot more than you'd think.)

If you can, grab yourself a whiteboard and some quiet to think things through. Ask yourself:?

1. What do you want generative AI to achieve? This should tie back to your business objectives. Make sure the tool you’re looking at has the capabilities to support your objectives today and can scale to support them in the future.??
2. Does your organization already have access to a generative AI tool? You might be surprised by how many software vendors have generative AI capabilities; it’s possible that one of your existing software vendors has already provided access to their generative AI technology or would be willing do so if asked. But remember, just because it’s already in place (or easy to put in place) doesn’t mean it will be the best fit for your goals. Conduct your due diligence on each individual tool.??
3. Will your organization only be using one tool? This is a big one, because it’s almost a guarantee that many of your team members are already using a public facing generative AI tool like Chat GPT. If you’re seriously considering a licensed tool for your business, will it still be acceptable for team members to use a free option? Under what circumstances, and for what type of material??What are the best practices around using a tool that has not been vetted by your ET team? Will everyone at your company only be allowed to use approved tools, and how will this policy be communicated and enforced?
4. How easily will it integrate with existing systems? Understanding what it will take to connect the AI to the systems you already have in place is essential. Your technologists and IT team will be an invaluable source of information here – they are perfectly positioned to understand what questions need to be asked and to scope any integration project, and they're going to give it to you straight when you ask them how much time it's going to take.
5. What type of support or maintenance does the tool come with? Ensuring that regular updates are part of the package is critical with any software, but especially important in a field like generative AI where improvement and innovation is a nearly daily occurrence. Is there exclusive training to help you upskill your?team? How regular are the updates, and do they happen automatically?
6. How flexible is the tool? Any tool you select should be customizable to your brand voice and should align with the needs and goals of your organization. This is one of the more subjective questions because every organization is going to value flexibility and customization differently, but it’s important to think about your organizational needs over the next several years as well as its needs today.?
7. Are the outputs high quality? Again, this may seem like an obvious consideration, but the field of generative AI is increasingly crowded. Do your due diligence to make sure that the tool you choose generates high-quality, human-like text that is coherent, contextually relevant, and free of grammatical errors. And know from the beginning that human beings are still going to be a massive part of your process. At this stage in the genAI game, I'd say the best tools can get you 30% of the way to a final written product at best. You need to upskill and engage your human team to get the rest of the way there.
8. What data security and privacy standards does it have? This is especially critical if you intend to use generative AI to polish or integrate on marketing plans or proprietary information. Understand where your data is stored, how it is or is not being used to train other models, and how licensing may impact those facts. Any tool you use should adhere to data protection regulations and respect user privacy. It should not store sensitive information unless it’s critical to the use case, and even then, only with appropriate consent protocols in place.?
9. Does it operate within ethical and legal standards? There’s a bold new frontier approaching around generative AI specifically and AI tools in general, all to do with ethics and regulation. Do your research and ask questions about how the model underpinning the tech was trained, how credit is being given to any copyrighted material that may have been used, and what plans are in place to make sure that the tool continues to operate ethically and legally in the future.??
10. Is close enough good enough? Generative AI is good for ideation and initial brainstorming; but in some cases, granularity and specificity are necessary for outcomes that meet brand standards.??
11. How much does it cost? Ah, the big question. Budgets are tightening across nearly all industries, so make sure you know what a professional license is going to cost, and how many seats you’re going to get for that price before you get too excited about any one option. Many tools have a free option and an elevated offering with more features and protections that they offer for a fee. Understand where your data will live and what it will be used for on any free tool – all tools have a cost, but you won’t always end up paying with money.??

Finally, consider: are you overstuffing the walrus?

We've all been in a position of having to learn new things quickly. When I was first starting out in my strategy career, one of my most common answers to 'do you know how to do ___________' was 'I can figure it out.' My spidey senses tell me that's where a lot of us are with generative AI. There's nothing wrong with being in the Figure It Out Phase - in some ways it's the best place to be, especially if you are brave enough to admit that you don't know what you don't know, and use the Figure It Out Phase to, well, figure it out.

Do you know exactly what you're trying to accomplish with AI, or are you making your best guess? Have you done this before or is this a brave new world? Have you seen a walrus, or are you extrapolating what one should look like based on other factors like lived experience and theory? There are no wrong answers to these questions but they do need answers. I can almost promise that there's going to be real temptation to overstuff the walrus just to get it over with. But, as Jim Carroll puts it:

...if you want to stuff a walrus you need both people who know their walruses and people who know their taxidermy.

That's key to this whole process. No one is an island, and you're not in this by yourself. The next critical step is building a team of people to help you get this done. As a nice side benefit your team building is a perfect time to ask if anyone, just by chance, has ever seen a walrus.