AI Marketing Campaigns

Lead Designer · 80% Lift in Campaign Usage

AI Marketing Campaigns for Auto Shops

I led the design of Tekmetric's first AI feature, helping independent auto shops run marketing they never had time for. Monthly campaigns sent rose 80%.

Role

Lead Product Designer

Timeline

2 months, 2026

Platform

Web

Team

PM, tech lead, 10 engineers

Stakeholders

Directors of product, engineering, and design

What I owned

The work end to end: framing the problem, running research, setting success metrics with my PM and tech lead, shaping the design strategy, and getting stakeholder buy-in.

Problems with Marketing

Why Campaigns Weren't Being Used

Shops bought the marketing add-on for booking, but only 40% used campaigns, the feature that keeps customers coming back.

Feature usage among add-on subscribers

Online booking

80%

Campaigns

40%

Shops bought the add-on to win new customers, not to bring them back.

A year after launch, Tekmetric's marketing add-on had 3,000 shops subscribed. Most came for acquisition tools like online booking. 45% of subscribers did not send a campaign in 6 months, and campaigns are what turn one-time customers into loyal ones. Without them, shops saw less value over time and the add-on risked churn.

‍Business goal: Reduce churn and move Tekmetric closer to an all-in-one shop management platform.

‍Success metrics: Campaign creation first, then appointments booked from campaigns, then repeat customers over time.

Metric

Baseline

Target

Campaign usage

40% within 6 months

60%

Churn risk

20%

Reduce 5%

NPS

Monthly appts booked

120 per shop

180+

Time to first campaign

30 mins per campaign

5 mins

Research: Identifying the product gap

"I know I should send something, I just don't know what."

Interviews with 20 shop owners showed the blank template was the blocker: Most had never done marketing and didn't know what to send. Most had no marketing background. Faced with a blank template, they didn't know what to say, so they sent nothing. Marketing mattered for the business, but it wasn't a priority they felt confident in.

Insight 01

Most shop owners don't have marketing experience

Insight 02

Values a tool to guide them on best practices

User persona

Rob M.

Owner-operator, independent auto repair shop

Runs his shop on Tekmetric

Pays for the marketing add-on

Bought it for online booking

No marketing background

“Marketing matters. But between my techs, my customers and the cars, it’s never today’s priority.”

Goals

Bring customers back for their next service

Grow without a marketing budget or agency

Stay in control of anything sent under his shop’s name

Pain points

A blank template leaves him with nothing to say

Stretched thin across staff, customers and repairs

On a busy day, marketing is the first thing dropped

Needs

Campaigns written for him, in his shop’s voice

The right customers at the right send time

Final say and a record of every change

Why he’s the fit

He already pays for marketing and trusts Tekmetric to make the call. He doesn’t want to learn marketing. He wants it done for him, with him in charge.

1st Iteration

Prewritten Templates

Prewritten templates lifted campaign usage about 10%, which proved shops trust Tekmetric's guidance but didn't move the number enough.
AI was on the radar, but I started with the simplest answer to the insight: give shops something to send. I tested 10 prewritten emails and text templates with a beta group of 10 shops over 2 weeks. Usage rose about 20%. That told me two things. Shops trust us to make the right marketing call, a value we had to protect. And templates alone weren't enough.

Templates test

40%

48%

Campaign usage, +20%. Better, but not the shift we needed.

2nd Iteration

The Pivot

Shop owners don't want to learn marketing. They want a tool that markets for them, with them in control. A second round of 10 interviews with shops showed owners are stretched thin managing staff, customers, and repairs.
On a busy day, marketing drops off the list entirely. The goal wasn't to make shops better marketers, it was to do the marketing for them under their supervision. That insight also gave a clear, validated reason to use AI.

Problem reframe

The goal wasn't to make shops better marketers, it was to do the marketing for them under their supervision.

Wireframing and Ideation

3 Directions of AI Campaign Building

1. Fully automated: AI sends campaigns on its own. Lowest effort, but it breaks the trust shops place in us and removes their control.

2. AI assist in the editor: The owner writes, AI suggests copy. Low risk, but still needs the owner to start.

3. AI drafts, owner approves (chosen): AI proposes ready-to-send campaigns, and theowner reviews, edits, or approves. It removes the blank page while keeping the owneras the decision maker.

The Solution

AI Campaign Builder

The builder learns the shop's voice, picks the audience and send time, and hands the owner a ready campaign to approve.
The AI campaign builder learns the shop's tone from previously sent messages, suggests campaigns for the right customer segment, and schedules them for the best send time. The owner stays in charge: every campaign is a draft until they review and approve it. To keep it familiar, I built on the existing campaign editor rather than introducing a new surface.

How a campaign gets made

01 · Data layer

Everything the shop already knows

Customer DB

Service history

Weather API

Calendar

↓

02 · Business rules

Deterministic rules spot opportunities in the data

Overdue oil

Winter tires due

Lapsed customers

High-value VIP

↓

15–20 opportunities

03 · Opportunity selection

LLM

Picks the 4–5 best opportunities for this week

Input Shop context, weather, recent campaigns, all opportunities

↓

4–5 selected

04 · Campaign generation

LLM

Writes full campaigns for the selected opportunities

Output Subject lines, email and SMS copy, CTAs, timing

Voice Trusted local mechanic, no marketing jargon

↓

05 · Shop owner

Human in the loop

Review and send

Edit if needed, approve, schedule the send.

Why two LLM stages?

Lower cost: the model only writes the 4–5 campaigns that matter

Business rules stay deterministic, outside the model

No wasted generation on campaigns that never send

Tradeoffs

My PM wanted to launch with an improved marketing dashboard showing AI's impact. I pushed to ship the data trail first, because shop owners rely on a record of every change in their shop, and adoption depended on trust. The dashboard followed as a second release.

Decision

What We Traded

Why

Default segments instead of AI-generated ones

Smarter targeting at launch

Cut build time and model cost; AI segments became a later phase

About 3 AI campaigns a week, not daily

More volume of campaigns

Matches marketing norms and keeps generation costs sustainable

Data trail before the impact dashboard

Showing AI's impact at launch

Owners adopt what they can audit; trust had to come first

Impact

Monthly campaigns sent rose 80%, and small shops got marketing they could never afford before. Within 6 months of launch, monthly campaigns sent rose 80%.
Against the goals we set upfront:
60% of AI drafts were approved without edits. The feature also set the principles for later AI work, including custom segments. The real win wasn't speed. Independent shops with no marketing budget now get quality campaigns at a fraction of what an agency ~$1,250 – $2,250/month would cost.

Metric

Target

Result

Campaign usage

40% to 60% in 6 months

80%

Monthly campaigns sent

Churn risk

Reduce 10%

20 > 18%

NPS

Appointments booked

180 appts per avg shop

200 appts

Time to first campaign

5 mins

30 mins to 5 mins

Monthly campaigns sent, indexed to the pre-launch average (= 100).

Learnings

What I learned: Shipping the simplest, low lift solution first, templates, was worth it. It proved trust before we asked shops to trust AI.
‍What I'd do differently: Test the AI concept earlier, or involve shops in setting voice from day one.
‍Next steps: AI-generated segments, the impact dashboard, send times tuned to each shop's own engagement data.