This B2B sales and gifting platform's own team had built the first version in-house. It couldn't scale, it was slow and buggy, login, payments, and security had problems, and parts of it were unfinished. Working inside their team of 14 over 30 weeks, I took over and rebuilt the core systems: campaign orchestration, personalization, a LinkedIn and Salesforce extension, contact management, landing pages, and the AI inside those flows. The platform has run around 10K campaigns for 20+ companies.
At a glance
The first version was built in-house by the client's own team. It couldn't handle the volume of campaigns, pages, and emails the business needed, it was slow and buggy, and login, payments, and security all had problems.
I joined their team of 14 and owned specific core systems end to end, from design through production.
A typical B2B revenue team finds prospects on LinkedIn, keeps records in Salesforce, writes emails in one tool, builds landing pages in another, and sends gifts through a third. Every handoff is copy and paste, and nobody sees the whole journey.
The platform's idea is one platform where a signal about a person becomes a personalized campaign. My job was building the systems that make those pieces work together.
Campaigns are stateful: the system knows where each recipient is and what should happen next. Long-running work, like generating thousands of personal pages or sending scheduled emails, runs in background queues, so the app stays fast for the person setting up the campaign.
Each step feeds the next. Contact data flows into campaigns, campaigns generate personalized pages, outreach triggers follow-ups, and a recipient's actions decide what happens next.
Spot a signal
On LinkedIn, in Salesforce, or in the contact list.
Decide the next step
Campaign logic and AI suggest what to do.
Personalize
A landing page and message made for that person.
Reach out
Emails, follow-ups, and scheduling.
Send a gift
The recipient picks a gift and confirms an address.
Measure
Opens, visits, responses, and gifts in one view.
Sales reps spend their day in LinkedIn and Salesforce, so the product goes there. The extension recognises LinkedIn profiles and Salesforce records, pulls out the person and company details, and offers the platform's actions right on the page. No copying data between tabs.
Each platform has its own adapter, so adding another source later doesn't mean rewriting the core workflow.
The AI does real work inside the product: generating landing experiences, helping set up campaigns, personalizing from contact data, and taking actions through the platform's own tools.
A campaign might start with an email and end with a gift on someone's desk. Gift selection, recommendations, and address collection are part of the campaign, so a click on a landing page can trigger a real-world delivery, and the result shows up in the same analytics.
~10K
campaigns run
200K+
emails and pages generated
100K+
gifts sent
20+
companies, 1,000+ users
1,000+
extension installs, 100+ active
14
people on the team
What happens next depends on what each recipient did. Tracking state per recipient makes real multi-step journeys possible.
Trade-off: More complex to build and test than one-off sends.
Campaigns can mean thousands of pages, emails, and jobs. Queues keep that off the request path, so the app stays responsive.
Trade-off: More infrastructure to run and monitor.
LinkedIn and Salesforce pages look and behave differently. Adapters keep that mess out of the shared workflow.
AI that can use the product's tools and data produces outcomes, not just suggestions.
Trade-off: It needs more context and careful limits on what it can do.
Yes, if the ownership is clear. On this platform I owned specific systems end to end, from design through production, inside a team of 14. What matters is a defined scope, direct access to decisions, and someone accountable for each system.
Put it inside tasks users already do, and give it the product's context and tools. AI that drafts a campaign from real contact data, or builds a landing page from a template, gets used. A generic chat box in the corner usually doesn't.
More work
2026
AI-generated demo rescued: 80% rewritten, now 100K+ messages a day
2025
Half-built AI-coded app rescued: 10,000 users and $40K staked
2026
AI-built farm rental platform rescued: 200+ customers
2026
AI-built personalization engine rescued: 100K+ pages generated