As a digital product manager and strategist, it is my responsibility to observe technological innovations and try to predict its effects on the future of software products. After analyzing the most hyped current AI product innovations as well as working on my own prototypes for AI-powered features and entirely new products, I can present a few observations and make a few educated guesses.

In this article I want to talk about these topics:

1. What are the main driving forces behind this innovation frenzy?
2. What are the key AI capabilities you need to know about?
3. How will software products change?
4. What could AI add to your product?
5. What capabilities will your company need?
6. What will your product managers need to know?

If you work in the software industry and have not yet been able to invest much time in analyzing the topic of AI and reflecting on its significance for your own company and career, this article should be of value to you.

Let’s dive right in!

What are the main driving forces behind this innovation frenzy?

AI has been around for a long time. Now the technology’s evolution has accelerated into a frenzy of innovations on all fronts. A few very important things have happened:

• Models are performing better and better on all sorts of benchmarks we have for measuring intelligence. In most categories, they already beat the average human by a wide margin, and are starting to outperform experts.
• AI has become much easier to use. It is no longer limited to specialists. It has become immediately useful to any user. Interactions are through chat or voice, not through code or complicated user interfaces.
• AI has become much easier to deploy. Not many companies have their own machine learning engineers and LLM specialists. OpenAI and its competitors provide very simple programming interfaces that allow any developer to add AI functionality to a product.

What are the key AI capabilities you need to know about?

This is a tricky one, because AI is just the umbrella term for an immense range of technologies. However, there really are so many advancements happening at the same time! So, if your only contact with AI has been ChatGPT, you should know that there are many more models with great capabilities:

• AI can understand text input in almost any language, no matter how poorly written it is. It even understands Swiss-German, a language that doesn’t have an official orthography.
• AI can understand and follow instructions, no matter how complicated and nested.
• AI can output text at any language level and in any desired form and structure, producing customised output for any user.
• AI can reason (although the experts seem divided on this one, but it’s only a matter of time either way).
• AI can visualize almost any sort of data, concept or situation.
• AI can turn speech into text and vice versa (with any voice it’s trained on).
• AI can interpret photos, diagrams, illustrations, comics and other graphics and provide detailed descriptions.
• AI can interpret and precisely describe videos.
• AI can produce videos, for example personal video avatars with correct lip movements and gestures.

And of course combinations of the above are also possible. Does that trigger something in the creative parts of your brain? Oh, and we're moving very quickly from long processing times to (almost) real time.

In my opinion, what we are experiencing is first and foremost a UX revolution. The way people interact with machines may be about to change radically. I may be getting too old to enjoy swiping and typing on my phone, so I'm looking forward to interacting with apps in a more effortless and productive way.

How will software products change?

With all these capabilities becoming much better quickly and more widely available, what will happen to the software landscape?

Let me try to make some hypotheses about how I think AI will change software products and what that might mean for your product strategy. It's hard to make predictions, but let's see what we can do:

1. Most of the time, users will not knowingly use AI and will not really care what powers beloved new features as long as they provide value. The feature “Summarize this text with AI” will become just “Summarize this text”. I think it will take a while for companies to realise that.
2. AI will make it much easier for users to tell the software what they want to do; and to achieve their goals. User interfaces are unlikely to change that much, but we will see some interesting new UX patterns evolve. Especially mobile apps may take radical paths.
3. Software will be able to communicate with its users in a much more tailored, personalized way. Some products will no longer use the same copy for all users, regardless of their age, technical affinity and level of expertise.
4. Most software products will offer some AI-powered features. Not necessarily because it makes sense, but because it has become so much easier to add AI to products. In most cases, it will add some value for the users, and in a few cases, it will actually improve business outcomes for the software vendor.
5. Technical and other support will be available 24/7, with no queues. It will be more helpful than before, and if it's not labelled, no one will know if they're talking to an AI or a human.

What could AI add to your product?

Before you get excited, here's a harsh prediction. As a result of easy access to sophisticated models, certain AI features will simply be expected by customers. They will not even be perceived as a bonus. Unfortunately, it will also be perceived as a shortcoming if your product lacks them. If AI makes sense for your type of product, some of your competitors will soon use AI cleverly and threaten your market position. True and meaningful differentiation, as always, will not be easy.

What makes sense for your product will depend extremely on its nature and target segments. Here is a list of possibilities I brainstormed.

• If user input is an important element, support it with meaningful suggestions and feedback.
• Enable your product to get to know the user and adapt to their needs.
• Provide easy access to information about your product, in a natural dialog. No more jumping off to browse help centers.
• Translate your UI and other communications into any language. It may not be perfect, but then neither is most human-written copy.
• Adapt your UI copy and other communications to the language and technical sophistication of your users. Use the right jargon for experts and simple terms for others.
• Provide helpful contextual information throughout the application. Context not only refers to the current UI page, but also to the data entered by the user and previous user actions.
• Identify intelligently when users or organisations would be better off with a different pricing model and help them switch.
• Personalize based on available personal information, account data as well as usage data: detect where in the user lifecycle a person is, and adapt accordingly detect admins, power users etc and talk differently to them than to simple users.
• Let users interact via chat or via speech instead of through input masks.
• Detect user problems proactively and offer help before the user has to ask for it.
• Create highly customized training sequences that enable users to reach their goals very quickly even in fairly complicated software.
• Provide excellent feedback channels that make the user feel heard, but also helps you get actually useable feedback.

You will notice that a lot of these things are already present in some of your favorite products. That’s because AI is not new; and many companies have built great AI teams in the past years - before ChatGPT made most people aware of the state of things in AI.

And I’ll admit: the functionalities I have listed are ambitious, and while AI is advancing rapidly, some of these capabilities might not be as refined or reliable in practice - yet.

Without a strategy, without a clear plan, without a systematic approach, companies risk implementing “something with AI” that is not worth the investment. If you don't have the expertise in-house: Seek advice!

What capabilities does your company need?

Well, it depends very much on your needs. Let's look at a few different situations.

• Some companies may be happy to use one of the big AI models without deploying anything on their own infrastructure. Running your own model, even a simple one, can quickly cost you more than you gain. Instead, all you need to plug into the API of one of the big models is some basic AI skills and developers. Some customization via configuration will adapt the model just enough to your needs.
• If you want to offer certain expected standard features with AI, but don't want to invest into in-house AI expertise, wait for product plug-ins that include these features (e.g. chatbots).
• For a more sophisticated but still simple approach, you can use one of the services that allow you to train a model on your own data. You'll need someone with a deeper understanding, but as these services evolve, the requirements won't be high.
• If you need AI to do something important for your value proposition, you need product managers with a deeper understanding of AI, not just how to use ChatGPT well. You need a bridge between technology and user needs.
• If you want AI to help you develop new differentiators for your business, you'll probably need amazing machine learning engineers in addition to expert product managers. Mediocre AI features won’t be differentiators.

The feasibility and cost-effectiveness of these approaches can obviously vary significantly between companies and industries.

I want to point out that most of these approches also require multidisciplinary teams that include UX designers, domain experts and more.

It is absolutely crucial that companies develop a clear AI strategy to determine the required capabilities for the future.

What will your product managers need to know?

Will product managers be able to just look at AI like a magic black box that can do anything and expect engineers to do the rest?

No, they will need to develop a very solid understanding of the different types of AI (that's an umbrella term for a vast range of models and applications). And they will need to understand the models they want to use in depth.

• Prompt implementation (strategy, writing, measuring/testing)
• Context management
• Economics (esp. unit economics, model training costs, operational costs, scaling)
• Analytics
• Security, hardening against prompt injection
• Quality assurance / management
• Service orchestration

It's definitely not too late for product managers to invest in AI skills. In my opinion, there will be a significant demand for people who understand AI beyond writing a few prompts in ChatGPT. Builders! As these roles can become quite technical, while business acumen is equally important, such talent will be in short supply.

‍

‍