Being an Instacart shopper is rewarding in several ways: additional income, flexible hours, and a lovely way to serve your community. Families depend on you to shop for them so they can spend more quality time with each other.

And using Instacart as a customer is incredibly helpful in that it saves a ton of time grocery shopping. Time that can be spent with family and friends, learning a new hobby, setting up for a party, or a million other things. 

But whether you’re a shopper or a customer, you’ll encounter issues and call the help center for assistance (which, face it, nobody likes to do, no matter how good the customer service team is).

Instacart, especially after the influx of users that joined the platform after COVID hit, began to receive an enormous number of calls to their help center.  

Shoppers and customers were reaching out so often that Instacart didn't have enough agents to meet the demand and still provide excellent service.

That’s why our small, but mighty team of three (1 design manager, 2 designers) built an end-to-end chat experience for shoppers and customers to get instant help from an AI assistant.

DESIGN PROCESS

AI ASSISTANT IN ACTION

SHOPPER CHATBOT

Here’s our shopper chat experience, available on mobile and mobile web only. Please note that I didn’t design the UI. We were restricted with how our UI could look due to the third-party platform we were using to engineer our workflows.

ACCESSIBILITY

Advocating for accessibility features was an uphill battle, and our team was only able to achieve minimal gains, like changing the chat colors to a brand-aligned, accessible green and an easily readable font. What I would’ve liked to see in-production include:

• multimodality, specifically voice capabilities that wouldn’t require a user to type

• more concise messaging (often we were limited by legal + content requirements)

• testing integration with screen readers, voice control, magnification, and other assistive technologies

  • labelling interactive elements

  • notifying the user when a new message arrives

  • using sounds when sending and receiving messages

  • including alt-text for images and videos

IMPACT

Since the deployment of the customer and shopper help chatbots in 2021, our team has saved the company more than $3 million alone through over 10 million user sessions.

INDIVIDUAL IMPACT

Building a large-scale AI solution is a big team effort (especially when you have a little team, ha), but it’s great to work on a small team because each person can see how large their independent impact is. 

Here are a couple of high-impact experiences I led and designed end-to-end:

Customer Use Case: Instacart Membership

Context

I noticed high cancellation rates for our Instacart membership program and wanted to find ways to save costs in membership refunds and decrease agent contacts.

Optimization

Due to low engineering resources, I made content-only changes to our membership workflow. I added additional information about membership benefits, gave customers as easy way to cancel, but did not offer the option for a refund until later in the flow so users don't immediately select that option.

Impact

Our self-service rate increased from 6% to 30%, meaning that more users went through this new flow. The changes to the flow also saved us an additional $1,500/week ($78,000/year) in customer experience costs by deflecting agent contacts, which equals to $0.05 in savings per global orders.

Shopper Use Case: Marqeta Card Declined

Context

Our highest reason for agent contact from our shoppers is when their Marqeta card declines at checkout. A Marqeta card is an Instacart-issued debit card that shoppers use to buy items. Sometimes the card gets declined at checkout for various reasons, causing shoppers to reach out, often stressed and frustrated.

Optimization

I shortened the flow and edited the content to be short and easy-to-understand for a more seamless and natural experience. I also worked with engineers to incorporate an API to authenticate the shopper's Marqeta card and introduced new variables to differentiate between shoppers' physical and digital cards.

Impact

Our self-service rate increased by ~4% and our agent cost savings increased by an additional $4922/month (~$60,000/year).

WAYS TO IMPROVE

• Improve accessibility features (see above '“ACCESSIBILITY” section)

• Make UI improvements

  • Make UI consistent across shopper and customer apps

  • Make UI of agent handoff page consistent with the rest of the chatbot experience

• Include voice capabilities for a multimodal experience

• Improve NLU capabilities to identify more utterances accurately

• Increase the number of automations

• Re-think priorities

  • Focus on optimizing and improving a few flows instead of focusing on covering the most use cases

  • Focus on added value the chatbot can bring rather than recreating experiences already in the app

• Create more personalized flows that remember user information (name, returning or first-time user, past orders, etc.)

• Introduce proactivity to suggest offers, help, or anything else to the user exactly when they need it