Working session between Mark Hope and Karly Oykhman (joined later by Sebastian Gant) covering three main areas: AI-assisted Apex trigger development in Salesforce for the [1] instance, a multi-pronged rescue plan for the at-risk [2] client account, and strategic refinement of Asymmetric's core value proposition.
Attendees: Mark Hope, Karly Oykhman, Sebastian Gant
Goal: Automatically prepend a timestamp and user initials to each new comment in the Salesforce Task object's comments field.
What was built:
- An Apex trigger deployed to the Task object that fires on save, prepending [date] [user initials]: [new text] above prior entries.
- A new "Edit Comments" quick action button added to the Task layout as a UX workaround.
Key technical constraints encountered:
- Salesforce Lightning strips leading newlines from text area fields — blank lines cannot be pre-positioned via Apex.
- Cursor position in Lightning text areas cannot be controlled server-side; it always lands at the end of existing content.
- The initial trigger treated the entire field value as new on each save, causing duplication. Fixed through iterative prompting.
Final state: Functional but imperfect. Users must manually place the cursor at the top of the field and create a blank line before typing. The pencil edit icon still needs to be removed.
Lesson: AI coding is iterative by nature — expect 30–45 minutes for tasks that look simple. Being more explicit upfront helps but doesn't eliminate the back-and-forth cycle.
Goal: Fix a modal that duplicated the "Opportunity Information" section when creating a new Opportunity.
Root cause identified: Lightning automatically pulls in every page section that contains a required field. Because required fields (Process_C, Loss_Reason) lived in the "Opportunity Information" section but were missing from the quick action, Lightning included the entire section twice.
Fix attempted: The AI agent added the missing required fields directly to the quick action. Fix was not visibly confirmed during the session — likely a caching issue. Needs verification.
Aviary's contract is at risk due to missed lead targets. Finance industry email targets use security bots that inflate open rates, obscuring real human engagement metrics.
| Track | Goal | Owner | Details |
|---|---|---|---|
| Nurture Campaign | Re-engage ~40k dormant contacts | Mark | Build HubSpot list (exclude active sales pipeline + ABM), send via Orbit/SES, set send limits |
| LinkedIn Outreach | Add high-engagement channel to ABM playbook | Mark | Manual outreach; request Aaron/Blessing's LinkedIn login; hire outreach resource |
| Google Ads Optimization | Improve ad performance and conversion | Sebastian + Karly | Fix final URLs, link business name/logo, add LinkedIn overlays, upload Credit Union customer match list, request +$500 budget for LinkedIn ad test |
| Website Rebuild | Improve site performance, enable faster landing page creation | Karly | Get time estimate from Eshak for WordPress rebuild |
Finance industry targets use email security bots that auto-open emails, inflating open rates and making it impossible to distinguish bot engagement from human engagement. This is a known deliverability/measurement challenge for this vertical.
The team refined the agency's positioning away from selling marketing services toward solving specific, high-stakes business problems.
Marketing services (SEO, ads, email, etc.) are the mechanism, not the value proposition.
Next step: Karly and Melissa to meet to refine Asymmetric's messaging based on this framing.
On AI-assisted development: Mark noted the dual experience of working with AI agents — simultaneously impressed and frustrated. The Salesforce session is a good example: the agent eventually solved the core problem but required significant iteration, hit platform limitations it couldn't work around, and occasionally made things worse before making them better. The value is still real — this work would have cost ~$10k with a Salesforce developer like Dimitri and taken days.
On Salesforce's trajectory: Both Mark and Karly observed that Salesforce is bloated and hasn't kept pace with how users actually want to work. The platform's complexity creates ongoing friction that AI agents can partially offset but not eliminate.
On AI agent safety for Google Ads: Mark cautioned Sebastian about invoking existing MCP tools without understanding what they do first. Audit tools may be configured to auto-fix issues, not just report them. Best practice: ask the agent to explain each tool before running it, and explicitly instruct it not to change anything during an audit pass.