Preface
If you’re someone who works in a SaaS (Software as a Service) product team, or in User Experience (UX), you might have spent some time thinking about what the serious advancements in AI mean for you and the products you work with.
As a firm that has worked for a decade with numerous companies (and governments) to build great products that are designed to effectively solve the problem they set out to solve, and engineer great user experiences, we have a lot to say on this topic. At the end of 2022, we started in a similar state of uncertainty on what AI means for us and our clients, but after 2.5 years and a significant amount of research and hands-on experience, we are ready to share these insights with you.
The TL;DR of it: this changes everything, and this changes nothing.
But that’s simplistic; allow me to elaborate over the course of this piece. Okay, just to be clear: we are not AI deniers. In fact, we’re the opposite of it—we are interested in deeply invested in helping our clients (and anyone, really) make thoughtful choices while implementing AI, and do it in a way that elevates user experience [See anyonecanai.io]. Hence the interest in what the entry of AI into every sphere of life and work could mean for established companies, new entrants, and most importantly: the users.
We have a lot to say about how AI impacts products and how they’re designed, but we need to start somewhere. So, in this particular deep-dive, we’ll aim to unpack how AI is disrupting web and mobile applications, what changes to expect, and how to be best-prepared for this new age of products.
Part 1: Evolution of UX in SaaS
SaaS is one of those words, which because of its ubiquity in our lives today, could be used to refer to way too many things that it has sort of lost its meaning. Try asking someone below 15 years of age what the World Wide Web is and you’ll know what I mean. Ask a tech journalist from the early 2000s about SaaS, and their eyes would light up as they tell you the story of its origins.
A path-breaking idea and concept at the time, SaaS startups that ventured outside the on-premise model of selling software, took up the largest share of Silicon Valley funding, and later, IPOs. But today, we live in a world where web apps are almost as common as websites, and we’ve taken SaaS products for granted.
What if I told you that this was made possible by two decades of UX work that went into it?
In the initial days, when offering software over the internet was a novel idea in itself, SaaS products had clunky interfaces and there was little importance given to user interaction1. Once the model gained popularity and a seemingly unending supply of funds flowed in, increasing competition, companies started thinking about ways to retain users; bringing focus to ease-of-use, and the interface.
Now, when applications were used over the internet, it became possible to collect more frequent—and hence more relevant—usage data and feedback. The conditions were ripe for the rise of UX as an important function in the building of products.
Logical user flows, essential buttons, and a somewhat pleasing interface became commonplace, encouraging (and simultaneously responding to) the mass adoption of these tools, especially those that were used by non-specialised users, such as Gmail, internet banking applications, etc. Then came the need for responsiveness to various screens, and then mobile-first designs.
By 2016, trends reports2 showed that on average, every dollar invested in UX brings 100 dollars in return, and that good UI can increase conversion rates by up to 200%, and good UX can double this—up to 400%!
It became clear that UX was here to stay.
Soon, the data collected on preferences and usage patterns led to personalization and intelligent recommendations—aren’t we all grateful to Netflix for allowing you to have a separate profile from our parents?
All this to say that while SaaS products might feel like monoliths that have remained the same since their origins, it is not true. They have evolved continuously, over the last two decades. And the advent of LLMs is just another advancement that these products have to reckon with—albeit a mighty one.
Part 2: What does AI mean for SaaS?
Speaking of monoliths that are most threatened by AI’s disruption, we must discuss B2B (Business to Business) Enterprise SaaS players. The most prominent among these operate in the space of CRM (Customer Relationship Management), HRMS (Human Resource Management System) and ERP (Enterprise Resource Planning): essential, high-volume workflows that large companies have to spend on and manage.
These tools are currently built to be managed by humans who are trained in their use—to input and structure data. They risk becoming costly or redundant, as AI is promising autonomous workflows in a few use-cases, and the ability to process and gather insights from unstructured data in others. So they’re faced with the urgent need to adopt AI themselves, as well as adapt their offering to the new reality.
To simplify, let’s take the example of customer service. Instead of call centres, and hundreds of humans speaking to customers to resolve complaints and then manually keying in details of the complaints and resolutions, now there are sophisticated AI agents in the form of chat bots that are trained for that specific use case: to collect and resolve as many complaints as possible, as well as autonomously analyse this unstructured data (in the form of text or images) to generate insights—very often, in real-time!
The use of such an AI agent might negate the need for a large portion of the services provided by the company’s existing CRM platform, pushing the latter to quickly adapt.
Wherein lies, at least a part of the problem.
The rush to adopt AI in an attempt to adapt
Ever since the end of 2022, when ChatGPT’s launch rattled the world (followed very quickly by the release of progressively more sophisticated specialised AI tools for purposes you didn’t even know existed3), companies have been under immense pressure to adopt AI. Fueled by a kind of corporate FOMO, and the fear of being displaced from their place in the market, what ensued was a game of blind-folded treasure hunt where a lot of funds were allotted to AI adoption and literacy programs. The result? Surface-level AI features now embellished unchanged products.
This crazy rush to adopt AI has been responsible for a trend of ‘sprinkling AI’ on a product that is fundamentally still the same—analogous to adding a few kernels of oats to a traditional white flour cookie and calling it an oat cookie.
While there is nothing wrong with this kind of a token AI feature in the short run, it is merely a band-aid solution, which can do little to protect the product from the impending disruption.
Reimagining SaaS products: AI-first workflows
So what is the sustainable alternative to adding an ineffectual chatbot or a surface-level “AI search” feature to save a SaaS product from disruption? Reimagining the product with AI-first workflows—the equivalent of simplifying the ingredient list and using cleaner alternative ingredients in the cookie example above.
Think: mother of all redesigns.
This requires going back to the drawing board. And without being constrained by outdated assumptions or current ways of doing, to reimagine the best and most efficient way to solve a problem, or provide an offering—in a way that makes most sense to the user—while tapping into the power of AI.
This would mean thinking beyond buttons and the Graphical User Interface (GUI), and accounting for speech or text prompts, and gestures; incorporating Natural Language Processing (NLP) into the back-end, opening up a whole range of possibilities with multi-modal interactions, including cross-lingual capabilities.
Additionally, this would mean effectively using all kinds of user data to create contextual interfaces based on the location and what is being asked, and providing personalised, intelligent recommendations.
Take it one step further, and the AI agent does not wait to be asked, but offers inputs and suggestions without being prompted (like your smartwatch urging you to get up and move after hours of sitting at a desk). As the last example illustrates, some of this is already happening.
Successful examples of AI integration in SaaS
Legacy SaaS giant Salesforce has successfully integrated AI and launched Salesforce Einstein4, which automates the updation of the CRM by syncing customer information, analyzing email interactions with them, and even flags incomplete entries.
Similarly, Zendesk, which operates in the same space, claims that 80% of customer interactions can be resolved by its AI chatbot, escalating only a small portion of requests that need human intervention to be resolved. In fact, the company recently announced5 an outcome-based pricing model where customers will only pay for successful query resolutions. This is a great example of a SaaS product completely reimagined with AI.
On the other hand, a new crop of AI-first companies are now building Vertical AI agents specialising in the automation of a very specific task, sometimes restricted to use in some of the most niche industries, and these kinds of solutions are emerging as the next big wave in tech, attracting investments in the Silicon Valley and beyond, or even getting acquired by older companies6.
The Vertical AI market was valued at $10.2 billion in 20247 and its market capitalization is projected to be at least 10x the size of legacy Vertical SaaS. These are the ones touted as killers of SaaS.
But it’s not as simple as that—we need to gaze deeper into the crystal ball to get close enough to what might happen. And we can do that by looking at the facts of the present, and lessons from the past.
The Real Impact of AI on SaaS
Superhuman, an AI-powered productivity-focused email app that is designed to save people time, is a brilliant example of a successful Vertical AI product. But in its current form, it works on top of Gmail and Outlook, and hence is not a direct threat to these apps, at least in the near future.
Typical of all new technological advancements—a lot of conversations on AI tend to be either heavy on hype, or leaning heavily towards fear-mongering. But if we stop grasping in the darkness, and calmly pay attention to what’s happening, it can get us closer to the truth.
Smart product designers today use AI tools that make them more productive–by giving them a starting point for wireframes, a rough draft for UX copies, inspiration for designs, etc. But they have an unshaken belief in the need for and the relevance of their skill-set for many years to come. It is important to note that this is not despite, but because of how AI lowers the barriers to entry, making it possible for almost anyone to build an app today.
Sounds counterintuitive, but it is based on a time-tested simple truth: when there’s an abundance of anything in mediocre-quality, a high-quality unique version of that thing stands out because it is rare, and hence more valuable.
If you recall, it is the same reason that led SaaS companies to start investing in UX back in the day. I guess history does repeat itself.
Part 3: The Relevance of UX
Soon after the launch of ChatGPT, the first profession that was disrupted was Writing, or Content Creation. It was the most obvious use-case, and marketing teams could now potentially churn out scores of articles in a day with just one person working with an AI tool, as opposed to having a whole team of writers on payroll. A lot of writing professionals may have gone through the five stages of grief over the next few months, until it started becoming clear that this new reality led to an even greater demand for high-quality human-centred content.
“As AI commoditizes software creation–making it easier to generate text, audio, video, and code–the competitive advantage shifts from the software itself to everything around it. Trust. Relationships. Domain expertise. The human stuff.”
Similarly, great user experience will emerge as the distinguishing factor for products that succeed in this AI race. And there is a big role for UX research and design in shaping the future of AI-powered products—yes, even for Vertical AI agents.
But in this new avatar, UX work might look a little different from what we’re used to.
UX in the time of AI
We’re entering what we like to call the Experience Economy, and let me walk you through what it is very likely to look like.
AI can make it possible to improve user experience in so many ways; with increased personalisation, human language processing, prediction and the ability to make decisions. With AI, the tool can actually predict next steps or queries, and answer or cater to them, making it more—wait for it—human-like!
‘AI-powered products’ do not necessarily mean something that will resemble ChatGPT, Gemini, or Perplexity. There is no one shape or format that these products need to fit. In fact, the challenge lies in thinking outside the proverbial box. This means that buttons and visual interfaces are not going to completely go away any time soon. However, every app might have a powerful virtual assistant, which would require integrating text and voice inputs into the design.
The new-age workflows and visual interfaces will be hyper-personalised and context-aware. For instance, a banking app that displays different features or options depending on whether the user is at home or in a store. Similarly, a fitness coaching app that tailors recommendations and tips based on the person’s motivations, schedule as well as barriers—thanks to its operating at the level of context ecosystems, spanning several connected apps and platforms.
The mandate of the UX designer will thus go beyond building simple, clean, and efficient designs; to building products that are exciting, enjoyable, and persuasive—adding emotional intelligence to the mix.
In other words, UX specialists and product teams will be expected to create smooth, thoughtful AI experiences.
Designing Good AI Experiences
Being freed from the shackles of restrictive user flows and the conventional GUI could be a dream-come-true for innovators and creative nonconformists; a rare opportunity to almost start from scratch. But that’s not all it is—the architects of these AI experiences also have the added responsibility of thinking about ethical use, privacy concerns, trust in AI, and more fundamental questions on what to retain for human action and what to automate.
In this honeymoon period of our tryst with AI, it can be easy to get carried away by its potential to seamlessly cross boundaries of apps and systems, understand and infer information in so many formats, and provide intelligent inputs, almost ✨magically ✨. But it’s crucial to look closer, ask questions, and raise guardrails where none exist.
Take the last example of the fitness app’s hyper-personalised recommendation on workout timings and type based on the user’s work schedule, sleep patterns, menstrual cycle, gym subscription details, personal commitments, and self-declared information regarding their emotional barriers to change. On paper, this app could be the perfect solution for a CEO-mom who struggles with finding time for workouts in her highly unpredictable days. But, implementing it would require far-reaching permissions to fetch and utilise data outside individual apps–the privacy concerns and ethical considerations involved in which we cannot even begin to discuss in this article.
As promised, without getting too caught-up in the details, we want to share a list of considerations for those who are designing AI experiences. We hope you find them helpful. If you have other ideas we seemed to have missed, write to us at hello@parallelhq.com. We’re eager to hear them.
1. Gather Input Data Effectively.
Design for good input data by making it easier for users to provide context. This reduces ambiguity, leading to more accurate results, faster processing, and lower costs. There are a few ways to do this. First, is by finding ways to get structured prompts. Most people still do not know what to ask and get from AI, so priming them with examples of effective prompts or queries (like Zoom AI) can go a long way. Similarly, Canva AI, showing a glimpse of the range of options, helps steer the user towards the desired output.
2. Keep Users Educated and Aware.
While the invisibility of the workings of AI in the backend give it that magical feel, it also contributes to doubts and skepticism, stemming from a lack of understanding. When people see AI as a mysterious black box, they will have difficulty trusting it.
If we want people to embrace AI, we must first build trust, starting with being transparent about what the AI is doing and why. For example, Perplexity's loading states provide real-time updates on the search process. If you’ve ever wondered why your Math teacher was obsessed with seeing all the steps involved in the solution leading up to the answer, you might have your answer now. You could also effectively use visual cues such as progress bars to indicate the hidden processes.
Follow Explainable AI methods to explain the AI’s decision-making, and the reasons why certain suggestions are being made. Netflix’s “Because you watched…” recommendations, and Google Ads’ “Why am I seeing this ad?” are examples of this.
And finally, just like any other true attempt at building trust, you must be honest and upfront about the flaws. So, keep users informed about the AI’s limitations.
3. Integrate AI into Workflows.
Rather than having a standalone AI feature or chatbot, integrate AI into the existing workflows of the app–something the users are already familiar with. Anybody who comfortably uses a browser would know how to use the ‘Find’ function. Instead of having them navigate to or click something in a corner, what if they could access the browser’s AI capabilities by simply typing hello into the Find function’s search bar? That’s what the Arc web browser has done.
4. Learn & Improve through Feedback Loops.
AI won’t always get things right–and this isn’t just because it’s still early days. At any time, AI models can become outdated or irrelevant if feedback isn’t gathered. So building a feedback loop is essential to ensure that the AI learns and improves over time. But you don’t need surveys for this. Intuitive methods like a simple thumbs up or thumbs down reaction added to the results, can help with improving reliability and reducing hallucinations. You might have seen this in ChatGPT. And gotten to experience the significant difference this kind of feedback makes if you’re an active user of the like and dislike functions on Netflix.
5. Aim to Augment, not Automate.
Lastly, remember to use this guiding principle while choosing where and how to use AI: use AI to enhance human abilities, not eliminate the need for human involvement. Implementing AI will never be as simple as ‘automate everything’ and ‘let the algorithm decide’, because we’re still building products that must createvalue for humans. So identify where human judgement and creativity are essential, and provide space for it.
“Humans change much less than the world around them. Our needs, desires and fears are remarkably consistent.”
AI has shown us some of the most astonishing and unprecedented changes in the way we live and work in recent history. And yes, the rate at which things are changing is dizzying (in fact, I know that things have already changed since I started writing the first paragraphs of this piece). But the truth is that the most important things—fundamental guiding principles behind what makes good products—will remain the same.
UX will be all the more important in this intermediate transitional stage when decision-makers, product teams and designers experiment with the possibilities. SaaS product teams and UX designers have a unique and exciting opportunity to be the pioneers of this defining moment in history.
While the specifics of UX work may start to look very different, the nature of what we do, and our allegiance to the user will anchor us as we chart a new path into this unexplored territory.
References
- "The Evolution of SaaS User Experience." Life Self Mastery, December 16, 2023.
- "The Six Steps For Justifying Better UX." Forrester, report RES117708.
- "Breaking new fragrance ground with artificial intelligence: IBM Research and Symrise are working together to develop new AI-powered perfumes." Symrise Newsroom.
- "AI Workflow Optimization for SaaS Products." Techtic
- "Zendesk debuts outcome-based pricing for AI agents." ComputerWeekly.com, June 14, 2024.
- "Thomson Reuters completes acquisition of Casetext Inc." Thomson Reuters Press Release, August 30, 2023.
- "Vertical AI Market." GM Insights.
check out these related blogs
.png)
