Digitizing Sales Engagement

Background

Prospective buyers and existing customers visit Autodesk websites expecting seamless, friction-free experiences. While sales pages listed phone numbers, support and education teams only offered chat or case-based help.

Problem

This set up an inconsistent experience leading to frustrated support seekers into sales channels.

Roughly 20,000 inbound calls hit the digital sales team each quarter - yet 44% of those were actually for post-purchase or educational support, creating unnecessary volume, wasting time, and leaving customers frustrated.

My Role

Lead UX Designer, AI/Conversation Designer

• Sole designer on this project

• AI Design Strategy, Competitor Analysis, Early Concepts, Wireframes,

• Managed a team of three designers working on other use cases for AI and initiated the first AI Component Library to promote centralized reusable UI patterns to streamline handoff and speed up implementation.

Initial Solution

NLP-based routing that routed inquiries to Sales, Education and Support teams.

• Worked with Data Science to understand the model and recommended using an initial topic selection plus collecting the user’s question to balance efficiency and risk

• Partnered with Content team to craft ~80 FAQs for the top use cases identified by Sales and Data Science

• Priority was to connect users to a sales agent via chat but keep phone number handy

• Designed flows to account for chats outside of business hours, unavailable chat agents, and false negatives from the model to ensure we do not turn away prospective buyers

• Worked with Data Analysts and Product Management to design an A/B test where 50% of visitors to the US site would see a chat link instead of sales phone numbers

ChatGPT happened 💥

We were gearing up to launch sales chat, when data scientists reported they had used ChatGPT to create an experimental model that generated answers to pre-sales questions. All teams hustled and leaned into the excitement to get it ready.

• Inject LLM + RAG-powered self-service into the chat.

• Quickly shifted scope so support automated question resolution

• Partnered with sales to identify top use cases

• Brought in conversation designer to shape model responses

• Collaborated with Legal and Data Science to set trust guardrails (e.g. error handling, turn limits)

Outcome

• AI-powered, digital-first sales chat experience that identified sales inquiries and gradually replaced phone numbers across the US

• 30% of phone volume successfully shifted to chat within 4 months

• 25% of chats converted to live sales conversations

• 3,127 chats completed, generating 1,056 qualified leads and $625K in billings

• Initial open rates were under 1% - added a time-based pop-up nudge that engagement by 15%

• Nudge success informed a broader nudging strategy across the customer lifecycle

• Education users were bypassing cues — implemented a hard redirect to maintain sales focus

• Component library reduced confusion and re-work across design engineering, and QA

Launching doesn’t mean they will come 🤔

A/B test was slow and traffic to the Autodesk Assistant remained low on other touch points as well. with were also not getting much traffic. To address this, I proposed and designed a nudging strategy that was applied across the platform to boost awareness and gently guide customers towards realizing the benefits of the Assistant.