SALESIQ · AI SALES ASSISTANT · SIMPSON STRONG-TIE
ABOUT THIS PROJECT
Simpson Strong-Tie’s territory reps sell on the road, but their answers live in five desktop tools. I set the design direction for SalesIQ, an AI assistant that follows a rep’s day from plan to follow-up, inside hard limits on legal risk, data, and trust.
MY ROLE
Lead Product Designer. Direction, scope and prioritization, research, and stakeholder alignment with the director of sales and the VP of Platform Engineering.
PRODUCT
B2B · AI sales assistant · Web app for desktop and phone
TIMELINE
Sept 2026 – rep pilot next
RESULT
4.0+ of 5 in executive review · In build, pilot next
The Challenge
Reps need answers on the spot, and every tool that has them is stuck on a desk.
GOAL
Set the direction for an AI assistant before build, in two weeks, and get leadership and sales leadership behind one version of it.
PROBLEM
Not a lack of data. Preparing one visit meant five desktop tools and 30 minutes, notes lived in the rep’s head, and reps were already patching the gaps with ChatGPT and Excel, which moved customer and pricing data outside company oversight.
CONSTRAINTS
Salesforce stays the system of record. No audio kept. Works with no signal. Every number defensible in front of a customer.
The Solution
Build the assistant around the rep’s day, not around a chat box.
The obvious move was a chat assistant: give reps a box and let them ask. I went the other way.
Reps don’t start the day with a question; they start with a route. So I organized SalesIQ around the loop they already work in and had the AI do the assembling at each stage. Research surfaced four stages, and the design uses three, Plan, Visit, and Follow up, because reps prepare on the road minutes before walking in, so preparation belongs to the visit rather than to its own step. Building it this way also answered both sides of the room: AI-native for leadership, built on the rep’s real workflow for sales.
Research
Three reps, three territories, four pain points in common.
I interviewed three territory reps across Quebec, Alberta, and a four-state US territory, walked the prototype through with the director of sales for Canada, and reviewed six tools on the market. What stood out was how much the three reps agreed. Four pain points came up in every interview, and the loop every rep described became the information architecture. The market scan added one thing: the voice tools that make logging effortless all keep the audio, which was off the table for us. That gap is where the transcribe-with-consent decision came from.
5+
tools a rep opens to prepare for one visit
30 min
to prepare one account in Salesforce and BI
2 wks
for an official planogram, long enough to lose the order
Product Principles
Five rules that decided what the AI does and what stays with the rep.
EVERY INSIGHT ENDS IN AN ACTION
Home ranks who to see and says why. Each insight has a next step, like Add to plan.
CAPTURE WITHOUT TYPING
One Transcribe button, after the customer agrees. The conversation becomes tagged notes; text only, no audio kept.
THE REP APPROVES EVERY WRITE
SalesIQ drafts Salesforce updates and shows the note each one came from. Nothing saves until the rep taps Save.
SHOW THE SOURCE
Every number carries where it came from and when, so a rep can defend it in front of a customer.
FIT WHERE THEY ALREADY WORK
Salesforce stays the system of record, Outlook gets the visits, managers approve prices in Salesforce, and today’s route works with no signal.
One loop: plan, visit, follow up
Plan from insights, not a blank page.
Home ranks the accounts that need a visit and says why. The director of sales set which insights show up and when, so the ranking follows business rules, not the model’s guess.
The week on a map.
New plan lays the days out around Outlook meetings; the rep drags stops around and confirms. There’s no text box on purpose. Reps who want to type can go to Ask.
One screen before the stop.
Today lists the stops in order. Each account opens with its SalesIQ insight, then open quotes, the last visit, and past notes, with Salesforce and SAP details one tap away.
Transcribed with consent.
The visit screen opens blank with one Transcribe button. The rep asks the customer first; then SalesIQ turns the conversation into tagged notes, text only, with no audio kept. Hands-free, so the rep can stay in the conversation.
A review that writes Salesforce.
When the visit ends, SalesIQ drafts the review: a summary with its consent record, then the quote, samples, order issues, and follow-ups it heard. The rep edits anything and saves to Salesforce in one tap, even offline. Approving every write adds a step, held to under two minutes, and that step is what keeps Salesforce clean.
One tap away, not the front door.
Ask answers questions about prices, stock, and competitor parts mid-conversation, with sources, and can draft a planogram from a quick setup. I kept chat secondary on purpose: reps needed answers they could defend, not a blank box to prompt.
The Outcome
Cleared review, in build, pilot next.
STAKEHOLDER IMPACT
A clickable demo of the full loop cleared the executive review, with every scored component at 4.0 or higher out of 5. We then ran validation testing with reps and prioritized features into a Beta and an MVP. SalesIQ is now in build, with a pilot in two Canadian territories next.
HOW WE’LL KNOW THE PILOT WORKED
I set the targets before build, so the team agreed on what good looks like. Three numbers, measured with two reps in the field:
30 → 5 min
to prepare for a visit, from 30 minutes to under 5
90%
of visits logged within 24 hours
70%+
of AI drafts saved without edits
THE NUMBER THAT MATTERS MOST
Drafts saved without edits. If reps keep correcting the AI, capture and tagging get fixed before anything else gets built.
Reps were already patching their day with ChatGPT and voice-to-text. The job was never convincing them to use AI. It was building AI they could trust in front of a customer.
How I ran it
One place for everything, from research to engineering handoff.
I wanted a more efficient way to run the project, so I built a project hub with Claude: one stop for the prototype, every screen and its notes, the research, prioritization, a roadmap, a snapshot of the Jira board, and open questions anyone can answer. The PM, the dev team, and I use it daily. It’s where decisions get written down with their trade-offs, where stakeholders catch up without a meeting, and where a dev starts a screen.
Prioritization sorted every feature into Must, Should, Could, and Not now, which is what let us call Beta done and start building. Eng handoff maps each part of the app to a Strong-Tie Design System component with its layout, behavior, and data, so engineering builds from the same source the prototype came from.
PRIORITIZATION
ENG HANDOFF
