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.
This Week
Perigon Signals
News Monitoring That Triggers Sales and Ops Automations
We talk about signals a lot in this newsletter: monitoring competitors, tracking market shifts, catching buying intent from prospect behavior, identifying at-risk customers before they churn. That's because the companies that act on information first win. The ones still relying on manual scans and gut instinct? They're always a step behind.
Most of the signals tools we cover focus inward. CRM signals track what your prospects and customers are doing. Product analytics flag usage drops. Lead scoring models watch email opens and page visits. These are all critical, but they only tell you what's happening inside your own ecosystem.
Perigon Signals watches everything happening outside it. It monitors over 200,000 sources in real time and uses AI to surface external developments that affect your business: competitor moves, regulatory changes, funding announcements, market shifts. This is the context layer most small and mid-size teams are missing entirely.
Key features worth knowing:
Context-aware alerts that analyze sentiment, relationships, and relevance instead of just matching keywords. Perigon's own testing found that Google Alerts missed 67% of relevant content entirely.
Delivers insights where you work via Slack, email, Salesforce, or internal dashboards. No extra tab to check.
Living datasets that update automatically as new information appears, connecting current events to past context so you see the full picture.
Early feedback on the ground:
Perigon is still a relatively young platform (launched on Product Hunt in early 2025, backed by $5M in seed funding from LiveOak Ventures). That means you're getting in early, with both the upsides and trade-offs that come with it.
Who this is best for: Operations leaders and marketing teams at SMBs who need competitive intelligence, brand monitoring, or risk detection but don't have a dedicated analyst on staff. Think of it as your always-on external research layer that feeds the automations and workflows you've already built.
Zapier's Free AI Assistants
Qualify Leads, Draft Follow-Ups, Triage Support Tickets, and More
Zapier's February 2026 update makes nearly 100 free AI agent templates available. Each agent comes preconfigured for a common workflow. Connect your apps, tweak the instructions, and it handles the rest. Just a few of the tasks they can support:
A sales agent logs prospect details in your CRM, posts a summary to Slack, and drafts a personalized follow-up email in Gmail.
A support agent reads incoming tickets, categorizes them by urgency, answers common questions from your FAQ, and escalates complex issues to a human.
A content agent researches trends, drafts social posts, and queues them for review in your publishing tool.
You pick a template, adjust the instructions in plain English ("skip the Google Sheets step; send updates to Slack instead"), connect your apps, and publish. Build time drops from hours to minutes.
On the backend, Zapier also upgraded to Claude Opus 4.6 (released February 5), which brings stronger multi-step reasoning and fewer errors on complex, chained automations. Zapier reports that 97% of its own teams now use AI daily to build and ship workflows.
What's working:
Fast setup, real results. One hands-on review found it took just 20 minutes to build a working lead qualification bot that routes tickets to Zendesk based on urgency. A separate Lindy.ai review reported building a sales call prep agent that pulled calendar info, researched clients, and formatted briefing notes into a Google Doc, replacing what previously took 30 minutes of manual work per call.
Non-technical teams can self-serve. SelectHub's 2026 review highlights that natural language instructions make agent creation accessible without engineering support, calling it "a top choice for non-technical users who need robust yet straightforward automation."
Template-first approach builds confidence. Zapier's own agents guide recommends starting from a template and expanding step by step. NisonCo founder Evan Nison used a sales outreach template to automate lead research, call prep, and CRM updates across his entire pipeline.
Enterprise momentum is real. A Zapier survey of 500+ enterprise leaders found that 72% are already using or testing AI agents, and 84% plan to increase AI agent investment in the next 12 months.
Limitations to know:
Activity-based pricing adds up. Free plans include 400 activities/month. Paid plans start at $33.33/month for 1,500 activities. One in-depth review noted that layering Agents, Chatbots, Tables, and Interfaces on top of your base plan gets expensive quickly with no unified AI bundle.
Accuracy sits around 80%. Zapier's own documentation frames agents as delivering roughly 80% accuracy, and recommends adding human review before high-stakes actions like sending payments, publishing content, or updating CRM records.
Complex AI tasks hit a ceiling. Multiple reviewers note that agents handle routine tasks well (summarizing emails, creating notes, processing leads) but fall short on sophisticated multi-modal or deeply adaptive AI work compared to dedicated platforms.
Debugging can be opaque. Capterra reviews from verified users flag that error messages lack specifics, and troubleshooting multi-step workflows sometimes requires technical knowledge.
Bottom line: If your team has been waiting for a low-risk way to start using AI agents, 100 free templates plus a smarter AI engine is a strong on-ramp.
Want help picking the right Zapier Agent for your workflow? Grab a free Discovery Session with our team. We'll map your biggest time sink to working automations.
Keep Your Best Sales Reps
Automation as a Retention Play
Your top sellers didn't sign up to spend their day updating CRM fields, logging call notes, and chasing approvals. But that's exactly what's happening.
According to Gallup's 2025 State of the Global Workplace report, 51% of employees are actively considering leaving their jobs. And in a UiPath survey of 6,400+ workers, 58% said automation could directly address their burnout.
Removing friction from the sales workflow is a retention strategy disguised as an efficiency one.
Start with the tasks your reps complain about most:
Auto-capture call notes and CRM updates using tools like Fathom or Fireflies.ai synced to HubSpot or Salesforce, so reps never manually log activity again
Automate deal-stage progression and follow-up sequences using Zapier or native CRM workflow triggers to move deals forward, enroll contacts in sequences, and assign tasks automatically
Route approvals and discount requests through Slack or Teams with automated notifications instead of email chains that stall deals
Use AI to draft personalized outreach by feeding call summaries and deal context into Claude or ChatGPT, giving reps a polished starting point instead of a blank screen
Reps who spend more time selling and less time on busywork stay longer and close more. Research from Microsoft and BCI found that teams using AI-powered automation saw a 68% increase in job satisfaction. Stop losing good people to bad processes.
Deloitte's $290K AI Mistake Was Preventable
New Research Shows How to Avoid the Same Mistake
Last year, Deloitte delivered a $290,000 report to the Australian government that was later found to contain fabricated court quotes, nonexistent academic references, and citations of researchers who never wrote the work attributed to them.
Nearly half of enterprise AI users have already made at least one major business decision based on fabricated AI output.
The problem runs deeper than occasional errors. User research expert Caitlin Sullivan tested the same customer data across Claude, ChatGPT, and Gemini over 100 times. Each model returned different themes, different confidence scores, and different supporting quotes. All presented with equal confidence. Teams using a single AI tool for analysis never see what their model gets wrong.
What to do about it:
Apply a business logic gut check. Before accepting AI insights, cross-reference them against what you already know from your CRM, sales data, or support tickets. If AI tells you price is the top churn driver but your CRM shows most cancellations happen after a feature removal, something is off.
Verify before you present. After any AI analysis, run a follow-up prompt asking the model to confirm each quote or data point exists in your source material. Flag anything it can't locate. As a faster alternative, randomly spot-check 10% of AI conclusions against your original data. Either method catches errors before they reach a stakeholder deck.
Cross-check with a second model. Run high-stakes analysis through two different AI tools. Fuel Cycle CSO Rick Kelly now runs what he calls "adversarial reviews between agents" to assess whether outputs pass a quality threshold before delivery. If Claude and ChatGPT give you different themes from the same data, that's a signal to dig deeper.
Brief your AI like you'd brief a new hire. Include the specific decision you're making, your business goal, and relevant industry context. Ethan Mollick at Wharton now says giving AI proper context matters more than any prompting technique.
Force code for any numbers. Tell AI to use Python for all calculations, groupings, and comparisons. Text-based math from an LLM is unreliable. Add one line to your prompt: "Use code for all calculations. Output the code and the final result."
Save your prompts and outputs. Keep a record of what you asked, what the AI returned, and what decisions those outputs informed. 76% of enterprises have already added human-in-the-loop review. A simple audit trail means you can trace any recommendation back to its source if questions come up later.
The handoff point is clear: AI can do the first pass on analysis. A human should own the final call on what goes in front of stakeholders.
Deloitte's errors sat on a government website for weeks before a single diligent reader caught them. Your stakeholder deck may not get the same courtesy.
SMART WORDS OF THE WEEK:
— Richard Branson
Tools like Zapier's AI agents and Fathom train your team to work smarter. Automating CRM updates, call logging, and approval chains treats them well enough to stay.
 |
Stay in touch with our team from anywhere.
