How to Make AI and Automation Reliable Parts of Your Toolkit

Nathan Weill here.

I'm the CEO @ Flow Digital, where we help companies unleash their full potential by strategically automating every inch of their workflows. Each week, I share hot AI and automation tips to help you move your business into the future successfully.

Google Pomelli

Turn your URL + Logo Into Automated, Branded Content

Small marketing teams waste hours jumping between Canva, Google Docs, and scheduling tools just to get a single campaign out the door. The handoffs between these tools cause brand drift, rushed captions, and inconsistent visuals.

Google Pomelli, a new AI experiment from Google Labs and DeepMind, aims to fix this by automating on-brand content creation from a single website URL.

Key Features:

• Business DNA profiling: Automatically scans your website to extract brand colors, fonts, tone of voice, and visual style.

• Campaign idea generation: Suggests targeted marketing angles based on your brand identity.

• Multi-format asset creation: Generates social posts, web banners, and ad creatives ready for download.

User Feedback

In hands-on testing, one digital marketing agency completed a full social media campaign in 30 minutes versus their usual 3-4 hours. Another reviewer reported first drafts appearing 30-40% faster than their typical Canva workflow. For SMBs without dedicated design teams, Pomelli provides brand consistency at zero cost during the public beta.

What's working:

• 95% brand accuracy match in multi-campaign testing.

• Zero manual brand kit setup required.

• 300 free image generations per month during beta.

Limitations:

• No direct publishing to social platforms (manual download required).

• Only one Business DNA profile per Google account.

• Limited creative diversity; visuals can feel template-based.

If you're spending more time wrestling with design tools than executing strategy, Pomelli's free beta is worth 30 minutes of experimentation.

Zapier: Auto-Enrich Every Lead, Contact, and Support Request

Sales and marketing teams waste hours manually researching leads, writing summaries, and categorizing feedback across their databases. Slate Magazine's branded content team faced this exact problem and needed a way to scale lead generation without adding headcount.

AI Enrich in Zapier Tables lets you auto-populate multiple fields in a record using a single prompt. The feature runs automatically when new records are created, generating structured information based on your existing data. No separate AI tools or complex Zap configurations required.

What you can do:

• Write a prompt once and apply it to every new record automatically.

• Reference any existing field in your table to personalize AI outputs.

• Choose from ready-made templates or build custom prompts from scratch.

Practical examples:

• Sales/GTM: A new lead submits a form with company name and job title. AI Enrich auto-generates a company summary, estimates company size, identifies likely pain points, and drafts a personalized outreach email.

• RevOps: Closed-won deals hit your table. AI Enrich analyzes deal notes and categorizes win reasons (pricing, features, timing) for pipeline reporting. This is similar to how teams use AI to run pipeline health checks 24/7.

• Construction: RFI submissions arrive from the field. AI Enrich extracts key details, flags urgency level, and suggests which subcontractor should respond.

• Customer Success: Support tickets sync to your table. AI Enrich summarizes the issue, detects sentiment, and recommends priority level based on account tier.

Slate generated over 2,000 qualified leads in one month using Zapier's AI capabilities for enrichment. If you're already using Zapier Tables, AI Enrich turns your database into an intelligent system that works while you sleep.

Ready to supercharge your Zapier workflows? Setup a Discovery Session with one of our certified Zapier wizards.

Say “Yes” to AI, with Rules

The Real Risk Employees Using AI Without Oversight

Teams defaulting to "no" on AI requests aren't reducing risk. They're creating it. Gartner predicts over 40% of agentic AI projects will be canceled by 2027 due to escalating costs, unclear business value, and inadequate risk controls. Meanwhile, employees are using AI anyway without oversight, generating what Greg Kihlstrom calls "confident nonsense": outputs that sound intelligent but need fact-checking. Without guardrails, teams mistake coherence for correctness and style for proof.

The solution: Replace the default "no" with a structured "yes, with rules" framework. As Gartner notes, the projects that survive focus on enterprise productivity with proper guardrails, not hype-driven experiments without oversight.

Build governance that enables speed:

• Approved tool list: Maintain vetted AI tools anyone can use without additional approval.

• Clear data red lines: Define which data types are off-limits (customer PII, financial records, proprietary code).

• SSO + role-based access: Tie AI tool access to your identity provider so permissions match job functions.

• Audit logging: Track usage for compliance, security review, and catching confident nonsense before it spreads.

• Fast exception path: Create a 48-hour turnaround for new tool requests.

The result: Organizations with mature governance frameworks deploy AI faster than those stuck in "no" mode. Clear policies remove uncertainty and position your team in the 60% of AI projects that succeed.

Claude Cowork and the New Automation Decision Framework

Anthropic launched Claude Cowork this month, bringing AI agent capabilities to non-technical users. Unlike a standard chat, Claude Cowork can autonomously plan and execute work directly on your computer.

Automation Platforms v. Cowork: When To Use What
Unlike previous automation tools that relied on brittle "if-then" logic, Claude Cowork uses visual and semantic reasoning.

Use Zapier/Make for high-volume recurring triggers, mission-critical workflows, and cloud app integrations with structured data.

Use Cowork for unstructured data (messy PDFs, inconsistent formats), local files without API connections, and ad-hoc tasks not worth building an automation for.

Pros & Cons
Early reviews are cautiously positive. In DataCamp's testing, Cowork deleted 27 duplicates and renamed files based on actual content DataCamp, handling tasks that would take hours manually.

Key limitations: The desktop app must remain open for sessions to continue, some projects require heavy token consumption, and macOS-only availability.

Essential Guidelines Before You Start
Cowork has direct access to files across your computer. Here’s how to build safely:

1. Create a Dedicated Working Folder
Never point Cowork at your entire Documents folder. Anthropic urged users to limit the tool's use to a single folder and keep separate backups.

2. Back Up Before Every Session
The bot could take "potentially destructive actions" such as deleting materials. Since there's always some chance Claude might misinterpret instructions, give very clear guidance and make sure your data is backed up.

3. Write Explicit Instructions
Vague prompts create problems. Specify exactly which file types to process, what "organize" means, and what should happen to duplicates. Including examples in prompts, such as "Rename files like this: report-2026-01.pdf," improves accuracy.

4. Limit Browser Permissions
Anthropic warned about prompt injection risks and advised limiting access to trusted sites when using Claude in Chrome.

5. Start Small
Test prompts in small scopes to verify behavior. Begin with a simple task like renaming a single file to understand the approval loop.

SMART WORDS OF THE WEEK:

— Kevin Kelly, co-founder of Wired magazine.

Building trust in AI tools requires consistent small wins with proper oversight. One careless deployment without guardrails can undo months of progress with your team and customers.

Stay in touch with our team from anywhere.