top of page
image.png

AI Assistant Upsell Experiences

Problem

Users attempting premium PDF actions (e.g., delete, rearrange, or rotate pages) within the AI Assistant received no guidance on how to complete those tasks. This created a dead end in the experience and missed an opportunity to convert high-intent users into trial starts.

 

Goal

Increase trial conversions by surfacing relevant PDF capabilities directly within the AI Assistant and positioning the assistant as an action-oriented workspace—not just a content consumption surface.

 

Solution

Designed and tested two AI Assistant upsell experiences triggered when users attempted premium PDF actions.

image.png

Variant A: Lightweight Prompt + Modal

  • Acknowledge the user's requested action.

  • Present a concise value proposition and free-trial CTA.

  • Show an informational upsell modal after click-through.

  • Route users to a verb-specific pricing page.

 

 

Variant B: Value-Led Response

  • Acknowledge the user's requested action.

  • Provide a detailed, bulleted explanation of premium capabilities.

  • Surface a free-trial CTA directly in the response.Send users directly to the pricing page, removing the modal step.

 

Test Design

Method: A/B Test

Groups: Control + 2 Variants

Language: English

Duration: Feb 2, 2026 – Mar 8, 2026

 

Control

No upsell guidance surfaced in the AI Assistant.

Success Metric

Increase free-trial starts and downstream subscription conversion from users expressing intent to perform premium PDF actions.

 

Rationale

The experiment was built around two key hypotheses:

  • Intent-based messaging converts better.

    Users requesting a specific PDF action are already demonstrating high purchase intent. Meeting that intent with contextual guidance should improve conversion.

  • The amount of information impacts conversion behavior.

    Comparing a lightweight prompt against a detailed value proposition helped determine whether users needed additional education before starting a trial.

Results

  • Projected annualized revenue impact: +$5.03M

  • Measured across English experiences in Reader and Reduced Mode.

  • The winning experience was shipped to production on March 20, 2026.

My Contribution

Led the content strategy and conversational UX for the AI Assistant experience, including:

  • Upsell messaging strategy

  • CTA and value proposition development

  • Experiment content design

  • User journey optimization from intent detection to trial activation

  • Performance analysis and recommendation of the winning experience

Trial Recap

1.png
2.png
3.png
4.png
5.png
6.png
7.png

Role: Content Strategist

Team: Acrobat Growth

 

The Problem

Trial users adopted Acrobat's tools, but the value built up during that usage stayed invisible to them. Usage happened in scattered moments. Value accumulated quietly. When the renewal decision arrived, users had no clear memory of what they'd gained, so the decision came down to gut feeling instead of evidence.

 

The Solution

We built a milestone-based recap that surfaces automatically during the trial and shows users a snapshot of what they've accomplished. Users qualify for a recap once they cross a usage threshold across tools used, sessions, or documents opened. This keeps the recap meaningful and prevents it from showing an empty summary to disengaged users.

 

Each recap shows documents opened, the most-used tool, and how many times tools were invoked. It also awards a status badge to power users, such as being in the top 20% for documents opened or the top 10% for edits, and flags when a user has returned to Acrobat in a later session.

 

Instead of a plain usage summary, each recap assigns the user a named archetype based on their strongest behavior, such as Document Explorer, Document Builder, Document Sealer, Document Pioneer, or Document Guardian. This turns a stats screen into something more personal and memorable.

 

The recap ends with a tool suggestion module. If the user shared an intent during onboarding and hasn't acted on it yet, we show intent-based suggestions. If they've already acted on that intent, we switch to machine-learning-driven suggestions based on actual behavior. If no intent was shared at all, the machine-learning suggestions are the default from the start.

 

Naming guardrails

The archetype name carries most of the emotional weight in the recap, so we set clear rules before naming any of them.

Focus on the archetype. Name the behavior or primary goal, not the person.

Keep it simple. Use two to three syllables.

Be accessible and inclusive. Names should be gender-neutral and understood globally.

Be clear and avoid stereotypes. Never imply how a person looks, their size, or their cognitive ability.

 

These rules mattered because a recap should make someone feel recognized for their effort. A poorly chosen name risks making them feel mischaracterized instead.

Outcome

Trial Recap turned an invisible pattern of product usage into a concrete moment users encounter mid-trial. It directly addresses the gap between quiet value and sudden decisions that were affecting conversion and retention.

8.png

Enterprise Onboarding Phase 2

1.png
2.png
3.png
4.png
5.png
7.png

Role: Content Strategist

Team: Adobe Design, B2B


Content decisions
 

Skip the reintroduction. For a migrated user, we did not write any copy asking them to set up PDF ownership again. Writing that step back in would have signaled that we did not know who this user was.

Lead with the payoff, not the feature list. The splash line reads "Meet the new Acrobat. Unlocking tools that elevate your work from day one." It names the outcome before naming any tool, since a returning user needs a reason to keep watching more than a rundown of what changed.

Remove the approval question before it gets asked. The overview card opens with "All AI-powered tools are approved for you to use by your admin." In enterprise products, permission is often the real reason a user hesitates to click. We wrote that line to close that question immediately instead of letting the user wonder and skip the tool.

Write toward action, not explanation. Each step in the three-part walkthrough is titled as an outcome the user gets, like "Get a quick summary" or "Chat with your file," and ends in a single clear call to action. We also pre-drafted the first AI Assistant prompt so the user's only decision is whether to hit send. The copy's job here was to shorten the distance between reading and doing.

Let repetition do the convincing. The overview card reappears once a week for four weeks. Since we knew most users would not act on the first pass, the copy needed to hold up on a second and third read without feeling repetitive or naggy.

Give every screen a way out. Every card and modal in the flow can be dismissed to Home. Writing a real exit into each step meant no one felt pushed through a tour they did not ask for.

Outcome

The content treated the migrated user as someone who already trusted the product and only needed a reason to trust the new parts of it. That framing shaped every line, from what we left out of the flow to what we chose to say first. 

6.png
8.png
9.png
bottom of page