This is a condensed recap of a Jukebox Podcast conversation between Nathan Wrigley and Matt Schwartz of Inspry, an Atlanta-based WordPress and WooCommerce agency (founded 2011). Matt also builds CheckView, a WordPress testing tool. The discussion explores how AI is changing agency work beyond design and copy: improving internal processes, client outcomes, and operational efficiency.
Background and perspective
– Matt came into web design early, kept building sites through school, and eventually founded an agency. Running a team forced him to document and refine processes he once managed alone.
– The focus here isn’t on AI generating pixels or headlines. It’s on using AI to streamline SOPs, client touchpoints, and delivery mechanics to create measurable value.
Client disclosure
– Matt’s rule of thumb: disclose when AI materially affects the client product. For purely internal automations that don’t change deliverables, disclosure is generally unnecessary unless you’re redesigning a client’s internal processes with AI.
Why adopt AI now?
– AI is accelerating capabilities and raising baseline expectations. Simple brochure sites and basic execution are being commoditized. Agencies that only execute risk losing competitiveness; those that emphasize strategy, process, and demonstrable business value will stand out.
– The gap is widening between agencies that adopt AI deliberately and those that don’t. Adoption should be intentional and aligned with your values and risk tolerance.
Create an AI vision document
– Matt recommends a written AI vision to guide adoption. Elements should include:
– A map of company processes and repetitive friction points.
– Decisions about where AI is appropriate, where humans must lead, and what guardrails are needed.
– Data-driven prioritization using time tracking, support logs, and process metrics—AI is best at recognizing patterns and highlighting automation targets.
– A documented vision sets expectations, protects reputation, and ensures consistent implementation across the team.
Productizing AI for clients
– Agencies can package AI-based services for clients who don’t want to manage AI themselves. Sell outcomes—automation, cost reduction, or revenue gains—rather than the tech buzzword.
– Useful offerings include lightweight internal web apps, automated workflows (e.g., n8n, Make), and bespoke automations tied to client processes. These can deliver immediate value and recurring revenue through maintenance and hosting.
– Caveat: don’t replace proven SaaS with custom builds unless you can maintain security, reliability, and support long-term.
Marketing and data analysis
– AI makes it affordable to aggregate and analyze large data sets for market research, persona identification, geographic targeting, and strategic recommendations. That lets agencies provide data-driven strategy to smaller clients at realistic prices.
Low-hanging operational uses of AI
– Meeting summaries and transcripts: AI tools can create concise notes, action items, and first drafts of follow-ups, speeding proposal preparation after discovery calls.
– Proposal and SOW drafting: AI can synthesize discovery notes, competitor sites, and past proposals to generate nuanced drafts that humans refine—an efficient middle ground between generic templates and manual writing.
– These uses increase throughput and responsiveness without proportional staffing increases.
Support workflows and caution
– AI can accelerate support by aggregating context—ticket history, logs, site state, server info—and suggesting likely resolutions for triage staff. This reduces escalations and speeds fixes.
– But support is sensitive. Over-automation can frustrate users, especially those needing accessible or empathetic service. Use AI to augment agents, not replace human judgment when stakes or emotions are high.
AI-assisted debugging and site management
– Tools connected to logs, APIs, and site data can surface probable root causes and suggest fixes, cutting hours from one-off debugging. Page-builder and CMS integrations are improving, enabling programmatic edits and agent-driven content changes.
– Matt advises retaining human review proportional to risk: have the agent share links and let a person verify before publishing.
– WordPress is increasingly an infrastructure layer for AI-driven management—AI proposes and executes edits while WordPress provides the capabilities—but human oversight remains essential until systems are consistently reliable.
Human-in-the-loop and trust
– Establish guardrails and review points. AI is powerful at pattern recognition and drafting, but unchecked use risks hallucinations, errors, or security lapses.
– Match oversight level to risk: high-risk tasks need strict human checks; low-risk tasks can be more automated.
Practical action items
– Prioritize process improvements where AI offers most leverage: discovery, proposals, meeting workflows, support triage, debugging, and routine automations.
– Build an AI vision document to define boundaries, responsibilities, and acceptable uses across sales, delivery, support, and development.
– Offer AI-powered services framed around business outcomes instead of technical explanations.
– Use AI to augment human expertise, not replace it—especially in high-friction or emotionally charged client interactions.
Next episode
– This is part one of a two-part series. Part two covers additional examples, ethical considerations, and more implementation details.
Credits
– Interview: Nathan Wrigley and Matt Schwartz on Jukebox Podcast (WP Tavern). For links and episode notes, see the WP Tavern podcast page.