โ† Back

Onboarding revamp: from hand-held to self-serve

Designed the onboarding journey where users experience the aha moment by getting their first RFP answered by SiftHub in under 10 minutes.

My role
Research to handoff + shipped some code
Collaborators
1 PM, 3 engineers
Timeline
3 weeks
Year
2026
Final onboarding: Preview step showing 12 RFP questions auto-answered from the user's connected sources
SiftHub

About SiftHub

SiftHub helps sales and pre-sales teams respond to RFPs faster using AI. It learns from a company's existing documents like past proposals, Drive folders, and SharePoint, and auto-generates sourced answers to new RFP questions.

Background

Onboarding was unguided, and the team was filling the gap

Most users sign up to SiftHub to try RFP filling. But the product wasn't guided, so they couldn't figure it out on their own.

They'd end up relying on our solutions team just to get started, making the core thing they came for dependent on a human conversation. That was a bad experience across the board, and it became unworkable as we expanded to startups where we couldn't dedicate someone to every new account.

No guided path

Users landed in the product with no sense of what to do next.

No way through without help

Even RFP filling, the core use case, needed a solutions team member to make it work for users.

Team was the workaround

Every successful onboarding required a human from our side. That wasn't sustainable at scale.

Current flow
Old onboarding: login and basic details before the AI teammate, with no guidance
AI teammate screen with no onboarding context or clear next step

Magic link โ†’ Basic details.

The challenge

Get to the aha moment in under 10 minutes

The biggest risk with onboarding: the moment a step feels like setup work, they bail.

The tricky part: users have to connect their knowledge sources before they can see any value. That's also the most likely drop-off point. The challenge was making a necessary friction point feel invisible.

Approach

Steps built around the aha moment

01

Profile

Pre-filled

02

Sources

Connect docs

03

Preview

Aha moment โœฆ

04

RFP

Your own RFP

01

Profile, we fill it in for you

After clicking the magic link in their welcome email, users land on a profile page that's already populated using information from the company website.

Profile, we fill it in for you โ€” step 01 screenshot
02

Sources, connect your knowledge base

This was the most critical step. Users had to give SiftHub their content for anything to work, but it was also the most likely drop-off point. The framing question shifted from "upload your files" to "what's the fastest way to give context to SiftHub?"

Sources, connect your knowledge base โ€” step 02 screenshot
03

Preview, the aha moment โœฆ

We pre-fill 12 sample RFP questions, so users don't have to upload their own RFP yet. SiftHub then generates answers in real time using their documents, complete with source citations. This is where users see the product's value for the first time.

Preview, the aha moment โœฆ โ€” step 03 screenshot
04

Upload RFP to fill

Upload or link a real RFP. Auto-detects columns, jumps to autofill.

Upload RFP to fill โ€” step 04 screenshot
05

Nudging users to install the Chrome extension

Most of our users prefer working in Google Sheets, filling RFPs directly there instead of in the web app. That workflow needs the Chrome extension. The challenge was nudging them to install it at the right moment, after they've felt the value, not before.

Nudging users to install the Chrome extension โ€” step 05 screenshot
06

Generating answers, educating along the way

Once users upload their RFP, SiftHub starts generating answers in the background (Web app flow). Instead of showing a blank loading state, we use this moment to introduce key parts of the RFP workflow, including sources, review, and export. By the time generation is complete, users are ready for the next step.

RFP questions and auto-generated answers
Source citations per answer
Review and refine answers
Export the completed RFP

Nudge to review answer.

Key decisions

What we chose, what we didn't, and why

01Profile step
What we did

AI pre-fill from company website

Company name, industry, competitors, and snapshot, all enriched automatically before the user sees the form.

What we considered

Blank form

Simpler to build, but it puts effort on users upfront and removes the personalization factor.

Decision 01 โ€” Profile step pre-filled with the user's company details
02Sources step
What we did

Skipped the website as a source

Focused on past RFPs and product support articles, along with sources like Google Drive and SharePoint that provide the depth needed for high-quality RFP answers.

What we considered

Using the company website as the starting point

Pulling information from the company website was the fastest and lowest-effort option since users wouldn't need to upload any content upfront. However, after reviewing past RFPs, we found that website content alone lacked the depth and specificity needed for high-quality answers.

03Preview step
What we did

Live demo on user's own data

12 RFP questions answered in real time from their connected sources, with citations.

What we considered

Sample data from popular companies like Stripe or ServiceNow

Faster to ship, but static and generic. The aha moment wouldn't feel personalised.

Decision 03 โ€” Preview step generating answers in real time from the user's own sources

Prototype

See it in action

Outcome

Quicker activation, shorter sales cycles

The redesign launched to both enterprise and SMB users. With users seeing value quickly and onboarding themselves without help, sales cycles got shorter and the team had more capacity to focus on bigger deals.

<6 min

Median time to aha moment

Most users saw their first AI-generated RFP answer (Preview) in under 6 minutes from sign-up.

70%+

Completed first RFP in session

Most new users went all the way from sign-up to a completed first RFP in their first session.

Self-serve unlocked. SMB customers could onboard independently for the first time, no solutions team required.

Support drop. CS team reported a noticeable drop in "how do I get started?" queries in the weeks after launch.

What's next

Expanding onboarding to cover every user type

This iteration focused on the first-time account owner. The next phase extends that thinking to two more scenarios currently in progress.

In progress

Invited user onboarding

When a team member joins an existing SiftHub account, their onboarding context is completely different โ€” sources may already be connected, company profile already set.

In progress

Use case based routing

Not every user comes in to fill RFPs. Some are there for deal briefs, email follow-ups, or collaterals. We're designing an early step that routes users into the onboarding path most relevant to them, so the aha moment is always specific to their job.

Up next

Keep exploring