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.
Descript AI Video Editor
Cut Editing Time by 70%
For teams producing webinars, podcasts, or training videos, the gap between "recorded" and "published" kills momentum.
Descript eliminates that bottleneck with text-based video editing and its AI co-editor, Underlord, which executes multi-step editing tasks from natural language instructions right inside the editor.
Key Features:
Text-based editing: Your main workspace is the transcript, not the timeline — cutting is as simple as deleting a sentence. Rimo
Underlord AI co-editor: Tell it to remove filler words, tighten pacing, and create social clips, and it constructs those changes automatically. It now supports an AI model picker including (Google Gemini, etc.) for different speed and accuracy tradeoffs.
Built-in generative media: Descript integrates video generation models like Veo 3.1, plus image generation with Flux 2 Pro so you can fill gaps in footage without leaving the editor.
Business Impact
Descript's text-based editing approach can cut editing time by an estimated 60–70% for spoken-word content. Recent updates also added a drive-level Media Library, so teams can upload and organize assets once and reuse them across projects.
What's working:
Non-destructive exports to Adobe Premiere, DaVinci Resolve, and Final Cut Pro make Descript a strong rough-cut tool that fits into existing professional workflows.
Underlord effectively eliminates the rough cut phase of production, handling audio cleanup, filler removal, and clip extraction in minutes.
Over 30 AI tools are bundled in — including Studio Sound, voice cloning, translation in 30+ languages, and AI avatars, replacing several standalone subscriptions.
Limitations:
No offline editing — Descript requires an internet connection for all operations which limits portability.
AI credit usage can burn through quickly and exceeding plan limits means purchasing top-ups.
Underlord's default caption templates can feel generic but you can build custom branded templates first to avoid AI-generated aesthetics.
If your team is still editing video on a timeline, Descript's free plan gives you enough runway to test whether text-based editing fits your workflow.
Pipedrive: Sequences Sell While You Sleep
Pulse Toolkit
Sales teams running Pipedrive are still manually deciding which leads to follow up with and writing individual emails for each one. Pipedrive's Spring 2026 release changes that with the Pulse toolkit, which pairs deal scoring with automated sequences so reps spend time closing, not chasing.
Scoring lets you define rules that increase or decrease a deal's likelihood to close Pipedrive, then Sequences let you build automated flows of emails and activities that pause when the contact replies.
What you can do:
Trigger sequences automatically based on deal stage, field values, or lead score, so hot leads get immediate follow-up and cold leads enter nurture tracks.
Send emails automatically or as drafts you manually validate, keeping a personal touch where it matters while automating the rest.
Route deals into different sequences based on engagement, deal size, or pipeline stage — no Zapier or Make required.
Great use cases we’re seeing:
Eliminate manual research: Fill in missing lead data to level up qualification and outreach with built-in data enrichment.
Reduce rep admin time by consolidating scoring, sequencing, and enrichment into one workspace instead of three separate tools.
Keep a human touch with personalized cadences while automating the timing and follow-up logic that reps forget.
Scale outreach capacity without adding headcount. Every rep gets an AI-prioritized task feed showing exactly who to contact next.
If you're on Pipedrive's Growth plan or higher and still building follow-up sequences manually, Pulse is the single biggest efficiency unlock in this release.
Scale Your AI & Avoid Costly Stumbles
Your Pre-Launch Checklist
The difference between companies that scale successfully (like Barclays, which ran a 15,000-employee pilot before expanding Copilot to 100K seats) and those that stumble (like Klarna, which had to rehire human agents after its AI chatbot created quality and trust issues) comes down to one thing: architecture before acceleration.
Ownership & Risk
If you turned off your main AI agent tomorrow, could you clearly map the operational and risk impact?
Who owns AI failures in your org today: a named role, or "whoever built the automation"?
When a Zap, Make scenario, or scheduled workflow breaks at 2 AM, who gets the alert? Who's checking the error logs in your SaaS tools on a regular cadence?
Visibility & Governance
Could a new hire understand how your automated processes work without calling the person who built them?
How many of your automations are documented and monitored vs. living in private accounts and one-off scripts?
Quality & Maintenance
Have you built proactive quality checks into your AI outputs: fact-checking layers, human review steps, confidence thresholds. Or are you just hoping the outputs are right?
When was the last time someone audited and updated the prompts powering your AI workflows? Models change, your business changes, and a prompt written six months ago may be producing quietly degraded results today.
Are your automations and AI tools working together as a system, or do you have a patchwork of disconnected workflows that no one has mapped end to end?
If those questions made you uncomfortable, you're not alone. A recent AI governance survey from Gradient Flow found organizations have models in production but most lag on fundamentals like monitoring, incident response, and staff training. One-off Zaps and agents solve local pain, but they create shadow systems that break under volume, process changes, or compliance scrutiny.
Treat 2026 as the year you move from experimentation to intentional automation strategy. Book a Discovery Session to find out whether your current stack is headed toward a Barclays outcome or a Klarna mishap.
Design Automation Just Leapt Forward
Point AI at Your Figma Files, Get a Working Page Back
The new Figma + Claude Code integration means you can hand Claude your existing Figma designs (brand components, colors, layout) and it writes code to produce a matching webpage inside Figma that your team can edit.
Figma struck parallel deals with Anthropic and OpenAI within one week of each other, signaling this is the new standard, not a niche experiment. Your design, marketing, and development teams now get:
Faster turnaround: Design-to-code translation that used to take hours now happens in seconds. Teams can go from approved mockup to working page in a fraction of the time.
Fewer revision cycles: Structural issues like steps that should be combined are easier to spot when laid out side by side on the canvas.
Revive old assets instantly: That legacy landing page no one has the original design files for? Capture it, pull it onto the canvas as editable layers, modernize it with your current design system, and push the updated version back to code.
Tips, Tricks & User Feedback
The richer your Figma library, the better the output: AI reads your existing design system (components, tokens, styles) and builds to spec. Teams with well-documented Figma files will see the biggest gains immediately.
It works both directions: agents can now write directly to Figma files, generating and modifying frames, components, and auto layout linked to your existing design system. Build in code, review in Figma, pull feedback back into the codebase.
One thing it won't do: scrape a competitor's website and auto-produce a matching layout. The AI reads your own files. To replicate an external site, feed Claude the live URL plus a description of what you want, and let it generate from there.
What users are saying:
One CTO's hands-on review found that roughly 85% of the output works impressively well, while the remaining 15% requires predictable, manageable human fixes.
Community developers are already building workarounds to automate repetitive Figma tasks—renaming, clicking, dragging—that have nothing to do with the actual craft of design.
Some users on Figma's forum report that setup doesn't always go smoothly, expect growing pains during beta.
Where humans stay essential: Claude gives you a beautiful starting point, but you still need to add state management, event handlers, and edge cases. AI handles the production grunt work; your team owns the quality bar. The 85% that works is genuinely impressive. The 15% that doesn't is where craft lives.
Figma's MCP is free during beta. Start with a single component—a button, a card, a form field—and try the workflow before scaling.
SMART WORDS OF THE WEEK:
— John Ruskin
AI can handle production grunt work, but the intelligent effort (governance, documentation, human review) is what determines whether your business scales or stalls.
 |
Stay in touch with our team from anywhere.
