Agentic Applications: Transforming Customer Journeys for Brands
This is the first blog in Yali Sassoon's Agentic AI series. Click the links below to access the other blogs:
Agentic applications, driven by large language models (LLMs), promise to reshape how consumers interact with brands. Rather than simply optimizing a single task, these artificial intelligence solutions can handle entire end-to-end workflows like managing your health and fitness or booking a holiday. In this post, we’ll explore what opportunities agentic technology presents brands for creating superior customer journeys, growing customer loyalty and driving competitive advantage.
Types of Agentic Applications in Today’s AI Landscape
To date, customer-facing agentic applications have not attracted much attention. To understand why, we need to first understand where they sit in the overall agentic application space.
Task or Persona-Centric Applications vs Workflow-Centric Applications
To date, most agentic applications are built to support an individual in a particular role (e.g. developers in the case of Cursor) or performing a particular task (e.g. perform a search with Perplexity). They may even replace the human agent altogether (e.g. Klarna’s AI Support Agents). You can think of these applications as being “task or persona-centric” - they help individuals or teams (a developer, a support team) perform existing functions more effectively i.e. better, cheaper or some combination of the two.
However, there are potentially bigger opportunities to re-imagine the broader value chains that those specific tasks form a part of. For example, the end-to-end process by which insurance companies receive, assess and pay out claims. There are a number of companies focused on solving just that e.g. EXL Insurance LLM by NVIDIA, Amplework Software and Sprout.ai amongst others. We can call these applications “workflow-centric”. Whilst most agentic applications today are vertical, there are even bigger opportunities to unlock value with workflow-centric applications.
Consumer-Facing vs Business-Facing Agentic AI applications
To date, most agentic applications help enterprises and the people working at enterprises to be more productive. Below are some examples collated by the team at Wing Ventures. You’ll notice that nearly all of these are departmental in the sense of being geared around a specific team persona or task, rather than an end-to-end process.
There are far fewer examples of technologies and solutions that are consumer-facing - my favorite two examples are OpenAI’s Operator and Perplexity. OpenAI’s Operator is especially interesting because it’s a “workflow-centric” application: it reimagines the end-to-end process for a consumer making a restaurant reservation, for example, rather than optimizing part of that experience (which might be searching for a suitable restaurant, using a vertical service like Perplexity).
Today there are more organizations focused on leveraging agentic applications in business-facing rather than consumer-facing use cases. This is because it’s much easier to commercialize those solutions – businesses will happily pay if you can show them an ROI. And there’s more focus on task or persona-centric applications than workflow-centric applications, because it’s easier to understand and solve for a task or persona than an end-to-end workflow. However, it is likely that there is just as much value in solving for broad customer-facing, workflow-centric use cases, and as I will argue below, huge opportunities for consumer brands to create value by building agentic applications.
Reimagining Customer Journeys with Agentic AI Applications
With these categories in mind, let’s explore a few real-world customer journeys to illustrate how agentic applications can eliminate friction, create rewarding new workflows and help brands stand out. Below are three common consumer journeys laden with complexity and time-consuming tasks, ripe for agentic re-imagination.
Feeding Your Family: AI-Powered Meal Planning and Shopping
I’m a father of four and there are few things I find as difficult as ensuring that we have a kitchen that is well stocked. Grocery delivery services help a lot but add more logistics to coordinate:
- Making sure that the groceries that are delivered are the groceries that we need (nothing essential should be left out, but we also want to ensure that there’s not too much excess or waste). In addition, we need to ensure that there is someone in to take the delivery at the assigned timeslot. This involves multiple steps:
- Regularly booking (ahead of time) delivery slots – leave it too late and the only ones that work are gone
- Keeping the basket topped up with new items as we work through the items that we’ve previously bought
- Planning the next few days - are there any school trips the kids will need packed lunches on? Are we having friends or family over for any meals?
- Ensuring that the basket has been checked out (payment made) far enough ahead of the reserved delivery slot.
- Figuring out what to make that’s tasty, nutritious, is liked by the whole family and makes good use of the ingredients currently available
- Actually making and serving the food!
There’s an opportunity here for an agentic application that:
- Observes patterns in a family’s meals
- Automates as much of the grocery ordering process as possible using real-time data analysis
- Makes the part of the workflow that can’t be automated a lot more intuitive and interactive - for example - inviting the customer to explain what is unusual about the next few days, so the agentic application can help the customer figure out what changes to make to the basket
- Suggests meals and recipes based on the tastes of family members and what ingredients they have available
Booking a Holiday: Multi-Step Travel Planning
Booking a holiday - especially an exciting trip to a country you’ve never been before, can be tough. If you want to make the most of a 1-2 week trip, there’s a lot of research that you might need to do to figure out:
- What places do you want to visit?
- What places do you want to stay?
- How are you going to travel between the different locations if you’re visiting multiple different places?
- What will you need to do ahead of time e.g. sort out any visas, get all the required vaccinations?
Then once you’ve done all that research you actually need to book the different items (flights, bus tickets, accommodation, restaurants etc.). This is not trivial as the same accommodation can be bought on multiple different websites: should you purchase directly, or through one of the many different aggregators? There are implications on cost and other factors e.g. cancellation. Then once you’ve figured out where to buy you have to actually make the purchase, being careful not to make any mistakes. Getting a date wrong can ruin a holiday, as anyone who has bought a flight for the wrong date can testify.
The two above stages are not actually separate - based on the cost of different options it is common to research alternatives and update itineraries to optimize for budget and availability, making the whole process even more complicated.
There’s an opportunity here for an agentic application that:
- Enables customers to describe, in high level terms, what sort of holiday they would like to book
- Provides a brochure-like inspiration experience, collating relevant information together for the customer so that they can start to assess the options
- Discusses with the customer how they feel about the different options presented, and use the information to support the customer narrowing down the set of options
- Provides pricing and logistics information related to each option
- Works with the customer to iterate to a single option and then make all the required bookings on behalf of the customer
Managing Your Health and Fitness: Personalized Wellness Through Data Integration
Managing your health is hard. It’s complicated and multi-faceted. It’s about how active you are - your cardiovascular health - but also your strength and flexibility. It’s about having a balanced diet. It’s about having enough sleep. It’s about your personal goals and your motivation levels. And the interplay of all the above.
Today there are a plethora of applications and devices for helping you with different aspects of your health and fitness. There are great apps to help you run a marathon within a certain time, do yoga a certain number of days a week or lift heavy weights. There are other apps to help you track what you eat and drink, and devices for measuring your sleep.
But even with all these different apps and devices, pulling all those different threads together is hard. And then when things come along to throw your regime off - travelling abroad or coming down with an illness - how do you re-adjust your regime?
There’s an opportunity for an agentic application that:
- Takes a much more holistic approach to a customer’s health
- Talks to the customer to understand their goals and health concerns
- Consumes all the different sources of relevant data – what the customer eats, how he or she sleeps, what exercise they do
- Develops a health plan for the customer, one that is iterated as time goes, based on the customer’s changing circumstances, preferences and progress against the plan. Crucially, the application should evolve the plan with the customer through a continuous feedback loop
Key Design Patterns for Re-Imagining Customer Journeys in Agentic Applications
We are still in the very early days of agentic applications - the speed and creativity of the innovation is inspiring. Already though we have identified two capabilities and associated design patterns that provide exciting new ways to help customers solve such problems:
Collaborative User Experience
On traditional (non-agentic) websites, the onus is on the customer to figure out exactly what he or she wants, and then perform all the actions required to reach that well specified goal. If an individual is researching a holiday, for example, she or he might search multiple different websites to explore different destinations, and then create itineraries for those different destinations. He or she is likely to consult official travel sites, blogs, social media and review sites as they make a decision. At the point that he or she makes a decision, he or she will then need to turn his or her attention to another set of sites and services to actually make the booking. The primary mode of interaction is clicking and scrolling. At best, websites can employ search boxes and personalization algorithms to save the user clicks and give them what they want faster. But the user has to work to identify and cross-reference all the relevant sources, and bounce between different services to iterate towards a preferred solution and then make the appropriate bookings.
Individuals can engage with agentic applications in much more flexible ways. They can talk to the agent (voice) or type open-ended questions. They can review the information provided (including clicking on links and scrolling content as before), but then actively give feedback - “I like this but I don’t like that - and this is why…” The agentic application can use that feedback to direct the next part of the user journey, building a better understanding of the customer and their goals.
Just as the user can engage with the agentic application more flexibly, the agentic application can respond to the user in a much more flexible way. It can pull multiple sources of information and condense it for the user into an easy to consume format. It can ask clarifying questions. It can make suggestions. And with each new piece of information that the customer responds with, it can better understand and progress the customer towards the customer’s goal.
The back and forth between an individual and an agentic system is a powerful interaction mode for a wide range of tasks, where the individual might not be 100% sure what she or he wants at the start and/or where an individual is looking to be educated as part of the journey. So in the case of managing your savings - as an individual learns more (from the agentic application) about the different sorts of investment product available, he or she might start to refine their savings goals. Or to return to the holiday booking example - the individual might start the journey with only a vague idea of what he or she wants, and this starts to form as he or she learns more about the different destinations, the attractions on offer, the history and more. In all these cases, the agent works with the customer to achieve his or her goals, rather than force all the cognitive load on the customer.
Removal of Thankless Tasks
Some tasks in the end-to-end customer journey are more enjoyable than others. Agentic applications can be used to remove the thankless tasks (by performing them on behalf of the customer), so individuals can focus on the enjoyable/rewarding steps.
In the case of booking a holiday - learning about the different destinations and attractions is a lot more fun than booking the different tickets required. Agentic applications are able to do those tedious tasks on behalf of the customer The agentic application should be able to check (and double check) that dates and times line up - so that a sufficient time is allowed to e.g. travel from the hotel to make a train connection - and recommend (or book on behalf) of the customer transport from the hotel to the station, for example.
In the case of feeding the family, many of the tasks are thankless. I’d love the agentic application to propose a shopping list to me, based on my recent purchases, and then let me modify it by explaining - “I’m having the extended family for dinner - so I’ll need more food - maybe you can suggest a nice meal for me to prepare? Something that works for my brother-in-law, who’s gluten intolerant? OK - so what extra ingredients will I need?” I’d like to be able to explain that for certain products I’m very open to substitution, but for others I’m brand loyal - and have it suggest substitutions (or if it’s confident, make them on my behalf.)
The Business Impact of Customer-Facing Agentic Applications
Consumer facing brands have an exciting opportunity ahead of them to create massively better ways for their customers to accomplish tasks - in other words, to differentiate based on user experience. Brands that are able to do this are likely to win new customers and better retain their existing customers.
This is a radical departure - over the last few years most websites and apps have standardized on the same design patterns - so that shopping on one website feels very similar to shopping on another.
As I will argue in a forthcoming post - this is more than just an opportunity: it is an imperative. Because if brands do not provide great agentic experiences for their customers, someone else will - and those companies' agents might end up consuming the brands website on behalf of their own customers, disintermediating them.
Agentic applications are poised to transform how consumers engage with brands. By tackling entire workflows end-to-end, they have the potential to deliver personalized, proactive, and truly collaborative experiences. As customer expectations evolve, brands that embrace agentic applications will stand apart, driving loyalty and growth.
Interested in building customer-facing agentic applications?
At Snowplow we’re building the data infrastructure to support enterprises building customer-facing agentic applications: applications that understand their customers based on all their behavior (what they’ve said, and what they’ve done, in real-time) - providing new capabilities and opportunities for example agents intervening in a customer journey if they see that a customer has got stuck or is in trouble.
Contact us to discuss how Snowplow’s customer data infrastructure can enable your brand to deploy customer-facing agentic experiences.
Credit to Pontus Noren, CEO and co-founder at Ensemble AI for this helpful categorization. Ensemble AI builds customer agentic applications for enterprises - if you’re at an Enterprise looking for support building agentic applications you should check them out.