The Top 3 AI Workflow Tools Transforming My Day

Updated May 2026. This is a substantial rewrite of a post originally published in July 2024. Two of the three tools have changed entirely. The take is the same: most AI is mid, a small amount is excellent at one thing, and that's where I spend my time.

When I first wrote this post in mid-2024, the AI conversation was about whether the bubble was about to pop. The Nasdaq had just shed a trillion dollars in a week, analysts were calling top, and half my LinkedIn feed was certain ChatGPT was a fad. The other half was certain it was the printing press.

Two years later, the answer is neither — and also both. The bubble didn't pop. Capital just kept stacking. Microsoft, Meta, and BlackRock have committed nearly $200 billion combined to AI data centers. But most of the AI tools people were excited about in 2024 are gone, pivoted, or quietly worse than they used to be. The hype rotated from chatbots to agents to MCPs to agentic workflows to "AI employees" — each one a 60-day news cycle, mostly a rebrand of something that already existed.

So here's where I've landed: AI is a tool category like any other. Some of it is genuinely useful. Most of it is mid. A small amount is excellent at exactly one thing — and that's where I spend my time.

This post is an update to the original. Of the three tools I wrote about in 2024, only one is still in my regular rotation. The other two slots have changed. Here's what's actually transforming my day in 2026, and what to skip.

What counts as an AI workflow tool, anyway?

Quick definitional setup, because the category has become a mess.

"AI workflow tools" used to mean Zapier or Power Automate with an AI step bolted on. Today, the term covers at least four distinct things:

Workflow automation platforms — Zapier, Make, n8n, Workato. Wire two apps together, optionally pass data through an LLM along the way.
AI agents — autonomous systems that take actions across multiple tools without a pre-defined trigger. Claude Code, OpenAI Operator, Lindy, Relay.
AI-native productivity tools — apps where AI is the product, not a feature. Cursor, Riverside, Granola.
AI features inside existing tools — Adobe Firefly inside Photoshop, Notion AI, HubSpot's Breeze.

This post is mostly about the last two categories. I'll cover the first two in a separate piece on workflow automation tools for B2B marketers — they're a different conversation.

If you've wasted too much time this week punching prompts into MidJourney or asking ChatGPT to do your taxes, the three tools below are the ones I keep coming back to. They have one thing in common: each does one thing extremely well.

1. HubSpot CLI + Claude Code

The setup that replaced my dev team

This is the most consequential change to my workflow in the last 18 months, full stop.

The HubSpot CLI lets you work on HubSpot themes, modules, and templates from your local machine. Pair it with Claude Code — Anthropic's terminal-based coding agent — and you get a development setup where you describe what you want in plain English and watch the files change.

How I Use It

I have built custom HubSpot modules for clients, rebuilt my entire website, and built two new apps using this stack. Things I genuinely could not have shipped myself a year ago. The workflow looks like this: open a project folder, describe the change, Claude Code edits the relevant files, runs commands, fixes its own errors, and pushes to HubSpot's design manager. I review, accept or revise, and ship.

For an under-resourced founder or a fractional CMO juggling multiple client environments, this changes the math on what's possible without hiring a dev shop. A few of the things I've shipped on this stack:

Custom HubSpot modules with structured fields and dynamic content
A full website rebuild with modern design tokens and animations
Two standalone web apps — one for content auditing, one for client reporting
Throwaway scripts for cleaning data, generating reports, and running bulk API operations

Why It's Effective

Real output, not slop: Claude Code writes production-quality code, follows HubSpot's module conventions, and respects existing patterns in the codebase. It's not generating "AI content" — it's generating working software.
Self-correcting: When something errors, it reads the error, finds the cause, and fixes it. No copy-paste cycle between an IDE and a chatbot.
Cost vs. agency math: A single HubSpot module from a dev shop runs $2,500 to $7,500. With this setup, the same module takes me a few hours and the cost is whatever Claude API tokens I burn — usually under $10.

Limitations

You still need to know what good looks like. Claude Code will happily build something that works but is structured badly. If you don't recognize a bad design pattern when you see one, it'll bake one in.
HubSpot-specific quirks need oversight. Field types, module schemas, and HubL templating have edge cases the model doesn't always anticipate. Test in a sandbox before pushing to a live theme.

If you want a deeper walkthrough of plugging Claude into HubSpot more broadly — content drafting, contact enrichment, and so on — I covered that workflow in Funnel Vision Issue 3.

2. Adobe Photoshop's Generative Tools

The one thing I won't replace with Canva

I work in Canva most of the time. It's faster, the templates are better for marketing assets, and my team can collaborate without anyone needing a $60/month Creative Cloud seat. For 90% of what I produce — social graphics, slide decks, newsletter headers — Canva is the right tool.

But there's one job Photoshop still does better than anything else: fixing a bad image.

How I Use It

Generative Expand, Generative Fill, and Photoshop's object-removal tools — all powered by Adobe Firefly — are in a different league from any of Canva's undercooked app library. If a client sends me a stock photo with a watermark, a product shot with a distracting bystander in the background, or an image that's the wrong aspect ratio for a hero section, Photoshop handles it cleanly.

The Adobe vs. Canva split in my workflow:

Canva: anything that starts as a template, anything where speed matters more than pixel-perfect output, anything a non-designer on the team needs to touch.
Photoshop: image cleanup, generative expand for hero crops, anything where the source asset needs serious repair before it can be used. The masking tools are still better in Photoshop too, but Canva is catching up fast.

Why It's Effective

Firefly's training shows. Adobe trained Firefly on their own licensed stock library, and you can tell — the generated content matches Adobe's general photographic sensibility. Expansions blend seamlessly. Object removal doesn't leave ghost outlines.
It integrates into the actual editing surface. Marquee a region, hit Generate, done. No round trip to a separate app, no prompt-engineering required for simple jobs.
It saves stock photos from the dustbin. Half the stock images I get from clients are cropped wrong for modern hero sections. Generative Expand turns a 4:3 image into a 16:9 in under a minute.

Limitations

Detail degrades with scale. Expand a small corner of a tree, fine. Expand a single cereal box into an entire grocery aisle, and you'll get hallucinated text on every label.
Licensing edges. If you're generating commercial content, double-check Adobe's current commercial-use terms for Firefly output — they've evolved since launch.
You're paying for Creative Cloud. The reason most teams default to Canva. If you only need image fixes occasionally, the standalone Photoshop plan is the most defensible spend.

3. Riverside:

The AI that made launching a podcast possible

I recently launched the Funnel Vision Podcast. It was something I'd wanted to do for over a year but kept putting off — the editing alone seemed like enough work to kill the project before it started.

Then I was invited to be a guest on someone else's podcast, and the host used Riverside. The guest experience was so smooth that I signed up for a free trial the next day. After about an hour of poking around, I realized: these are the best internal AI tools I've seen in any consumer-facing product.

How I Use It

For each episode, Riverside handles:

Filler word and pause removal: one button, and every "um," "uh," and three-second pause disappears. The cuts are imperceptible — no audio artifacts, no choppy transitions.
Audio enhancement: cleans up grainy or echoey recordings automatically. Guests who recorded from a coffee shop on AirPods come out sounding like they were in a booth.
Auto-resize for social: vertical clips for Reels, square clips for LinkedIn, full episodes for YouTube. Captions auto-generated. Speaker tracking happens in real time.
Transcription: accurate enough to use as the basis for show notes and SEO-friendly episode pages without major editing.

What's notable is that Riverside doesn't try to do everything. It's not a CRM. It's not a content calendar. It's not "an AI agent for podcasting." It records, cleans, and exports — and it does each of those things excellently.

Why It's Effective

Scope discipline. Most AI-native tools chase feature parity with adjacent categories and end up mediocre at everything. Riverside picked a lane and got world-class at it.
It removes the single biggest podcast-launch barrier. For solo operators or small teams, post-production is what kills the project. Riverside reduces a 4-hour edit to 20 minutes.
Guest UX is part of the product. Guests record on their own browser, no software install, no audio setup. The host gets clean local recordings of every participant.

Limitations

It's expensive at the higher tiers. The free trial is generous; the paid plans get steep quickly if you're publishing weekly.
Not a creative tool. Riverside cleans recordings — it doesn't write your show, choose your topics, or book your guests. If you're hoping for "AI does my podcast for me," look elsewhere (and good luck).

What about AI agents?

Half the comments on the original version of this post asked some variation of "what about agents?" — and back in 2024, the honest answer was "they don't really work yet."

That's changed. Agents work now, in narrow domains, when scoped well. I've written about a few that actually deliver:

Chatbase for customer support agents that can perform actual actions (refunds, order lookups) instead of just answering questions.
Zapier MCP as the "plug" that lets AI tools talk to 8,000+ apps using plain-language instructions.
Notion + GitHub agents for keeping documentation in sync with shipping code.

The pattern across the agentic tools that actually work in 2026 is the same as the pattern across the productivity tools in this post: narrow scope, one job, done well. The "AI employee that does everything" pitch is still mostly vapor. The "AI that does one specific thing inside an existing workflow" pitch is delivering real value.

The Bigger Picture

So what did the 2024 post still get right? The skepticism, mostly. Specifically:

Beware tools that try to do everything. The AI products that have aged best are the ones with the narrowest scope. Riverside is THE podcasting platform because they owned the niche and kept their scope dedicated to that customer base.
Hype cycles are not product cycles. The tool getting the LinkedIn buzz this month is almost never the tool you'll be using a year from now. If something is worth using, it'll still be worth using in six months.
AI doesn't replace judgment. Every tool above requires me to recognize when the output is wrong or the platform is hallucinating. Many on-board AI tools build in human decision making now. But if you've, for example, set Claude Code on a task and haven't restricted it, it can go down many blind alleys and waste your time.

The three tools above earn their place in my stack because they save me real hours, do one job extremely well, and have stayed useful over time.

Frequently Asked Questions

Which AI workflow tool is best for B2B marketers in 2026?

Depends entirely on what you're trying to do. For development and technical work inside your marketing stack, Claude Code paired with the HubSpot CLI is the highest-leverage tool I've found. For creative production, Adobe's Firefly-powered tools are still the best for image cleanup, while Canva is the right default for everything else. For content production workflows, Riverside is the standout for audio and video. There's no single "best AI tool" — the right answer is matching a tool with a narrow purpose to a specific job you actually do.

Is ChatGPT still useful as a workflow tool?

Yes, but the role has changed. In 2024, ChatGPT was the default for almost any AI task. In 2026, it competes with Claude, Gemini, and a dozen specialized tools — and for most workflow use cases, one of those specialized tools beats it. ChatGPT is still excellent for ad-hoc reasoning, drafting, and quick research. It's less essential as a workflow integration than it used to be.

What's the difference between AI workflow tools and AI agents?

An AI workflow tool helps you do a specific task faster — generating an image, transcribing audio, drafting an email. You're still in the driver's seat. An AI agent takes actions on your behalf across multiple tools, often without step-by-step instructions. You give it a goal; it figures out the steps. Agents are powerful in narrow domains but still require careful scoping. If you're using AI to help you do your job, you want workflow tools. If you're using AI to take a job off your plate, you want an agent.

Can AI replace a marketing team?

No. AI replaces specific tasks within a marketing function, not the function itself. The teams getting the most value out of AI in 2026 are the ones using it to eliminate the busywork that used to consume their best people — so those people can spend more time on strategy, customer conversations, and creative work that AI is still bad at. If you're trying to replace your marketing team with AI, you don't understand what your marketing team does.

How much should a small B2B team spend on AI tools?

Less than you'd think. The three tools in this post total roughly $80 to $200 a month, depending on tier — Claude API usage scales with actual work, Adobe Creative Cloud is around $60/month, Riverside ranges from free to about $40 monthly. The bigger cost is time spent evaluating and switching tools. Pick a small number of products that do one thing well, learn them deeply, and stop chasing the next launch.

Want help building this into your business?

I help B2B startups and growth-stage teams set up the kind of AI-powered automation and analytics workflows covered in this post. If you're trying to figure out which tools are worth your team's time — and which are noise — let's talk.

For weekly walkthroughs of specific AI workflows worth setting up, the Funnel Vision newsletter ships every Thursday.