What 30 years of digital marketing taught me about separating the signal from the noise in AI adoption.
Let me start with a confession: I’ve been doing this digital marketing thing for over thirty years, and I’ve never seen anything quite like the current AI tools explosion.
Every week, someone launches a new platform that promises to revolutionise your workflow, automate your life, and possibly make you a cup of tea while it’s at it. Most of them are rubbish. Some of them are genuinely useful. And a precious few have fundamentally changed how I run three agencies across email marketing, digital presence, and B2B performance.
Here’s the thing that nobody wants to admit, though: 72-88% of organisations now use AI in at least one business function, according to McKinsey’s 2024 research. Generative AI usage doubled from 33% to 65% in just twelve months. Yet (and this is the bit that should make you spit out your coffee), over 80% of organisations report no tangible enterprise-level impact from their AI investments. That’s a lot of money being spent on tools that aren’t moving the needle.
So why does this gap exist? Because most people are buying AI tools like they’re collecting Pokémon cards. Gotta catch ’em all! Except you don’t. 🤷♂️
You need the ones that actually work for your specific situation. And that requires understanding both the data and the practical realities of implementation.
I’m going to share what the research says about AI adoption across different professions, what the ROI data actually shows (spoiler: it’s more nuanced than the LinkedIn gurus would have you believe), and then I’ll walk you through my personal stack. The tools I actually use every day to operationalise and scale my portfolio across sendXmail, ZOPPLY, and HOT Leads.
AI Tools TODAY
The State of Play: Who’s Actually Using AI and Why
Before we dive into tools, let’s look at what the numbers tell us. Because if you’re going to make sensible decisions about AI investment, you need to understand where the technology is genuinely delivering value, and where it’s mostly hype with a nice logo.
Marketing: The Early Adopters (And Occasional Over-Enthusiasts)
Marketing professionals lead the adoption charge with 85-88% adoption rates (source). This makes sense, since we’re the ones who have to produce endless content, analyse campaigns, and somehow convince the C-suite that our latest LinkedIn post was worth the investment. CoSchedule’s 2025 research found that 79% of marketers report increased efficiency, saving approximately three hours per content piece and 2.5 hours daily overall.
Now, before you get too excited, let me be clear: saving three hours means nothing if you’re producing content that nobody reads. The productivity gain is real, but it needs to be paired with strategic thinking. AI can help you write faster; it cannot help you think better. That’s still your job.
Legal: The Most Dramatic Acceleration (From Sceptics to Converts)
Here’s a plot twist: lawyers (yes, the same profession known for fax machines and paper filing systems) showed the most dramatic acceleration in AI adoption. According to Clio’s Legal Trends Report, adoption jumped from 19% in 2023 to 79% in 2024. That’s a four-fold increase in a single year. The ABA’s more conservative measurement showed attorneys using AI-based tools grew from 11% to 30%.
Two-thirds of legal AI adopters now save 1-10 hours weekly, with three-quarters of hourly billable tasks potentially exposed to automation. When lawyers start adopting technology this quickly, you know something significant is happening. These are people who still use Latin phrases unironically. 😅
Software Development: Nearly Universal Adoption
Software development has achieved near-universal adoption: 97.5% of software companies have integrated AI into development processes, according to Techreviewer’s 2025 survey. GitHub Copilot now reaches 15 million users across 50,000+ organisations, with 90% of Fortune 100 companies deploying it.
The productivity data here is among the most robust: 55% faster coding with Copilot, and 82% of companies reporting productivity gains exceeding 20%. This is probably the clearest success story in AI adoption. Developers have found genuine, measurable value in AI coding assistants.
Finance, HR, and the Rest of the Pack
Finance functions experienced a 21 percentage-point jump in AI adoption (from 37% to 58% between 2023 and 2024), according to Gartner’s CFO survey. KPMG found that 92% of companies report AI initiatives meeting or exceeding ROI expectations in financial applications. The use cases making the biggest impact are intelligent process automation (44% of finance functions) and anomaly detection (39%).
Human resources climbed from 58% to 72% adoption between 2024 and 2025. SHRM’s research shows organisational HR AI use nearly doubling from 26% to 43%. The productivity signal is strong: 63% of HR leaders report productivity boosts, with resume screening and AI-generated job descriptions driving the fastest adoption.
Sales and business development lag behind marketing despite sharing many use cases, with 43% adoption compared to marketing’s 85%+. HubSpot’s research attributes this gap partly to trust issues around AI-generated customer communications. Fair enough. Nobody wants to receive a sales email that reads like it was written by a robot having an existential crisis.
The ROI Reality Check: Why Most Companies Aren’t Seeing Results
Here’s where things get interesting.
The most important insight from ROI research is the disconnect between individual productivity gains and enterprise-level value capture. At the task level, the gains are substantial and well-documented: 60%+ reduction in task completion time according to Stanford/World Bank research, 22.6% average productivity improvement per Gartner’s survey, and role-specific gains ranging from 13.8% (customer service) to a whopping 126% for developers in controlled studies.
Yet macro-level impact remains modest. 🤷♂️
Federal Reserve analysis found AI has contributed roughly 1.1% aggregate productivity increase from current adoption levels. That’s significant at scale, but far below individual task improvements. Penn Wharton’s Budget Model projects 1.5% GDP increase by 2035 and 3% by 2055. Value realisation happens gradually, not through immediate transformation.
So why the gap? 🤔
BCG’s research provides the most actionable answer: 70% of AI implementation challenges stem from people and process issues, 20% from technology problems, and only 10% from AI algorithms themselves. This inverts where most organisations allocate resources.
The failure rate data is sobering: 80%+ of AI projects fail according to RAND Corporation research. Gartner predicts 30%+ of generative AI projects will be abandoned after proof of concept by the end of 2025 due to poor data quality, inadequate risk controls, escalating costs, or unclear business value. S&P Global found AI project abandonment rose from 17% in 2024 to 42% in 2025. 😬
That brings the question: What separates the winners from the rest?
BCG found that high-performing organisations follow what they call the 10-20-70 rule: 10% of resources on algorithms, 20% on technology and data, and 70% on people and processes. They focus on fewer, higher-priority opportunities (approximately half as many initiatives as lagging peers) and embed AI directly into business processes with KPI tracking.
Here’s a pattern worth noting: purchased AI tools succeed at approximately 67% according to MIT research, compared to just 22% for internally built solutions. Menlo Ventures found that 76% of AI solutions are now purchased versus built, up from 53% in 2024. Ready-made solutions reach production faster and with lower risk.
Unless you have a very specific need and serious engineering resources, buying beats building. Well, in most cases, even though I have found a different trend. 😉
My AI Stack: What I Actually Use to Run Three Agencies
Right, enough theory. 😉
Let me walk you through the tools I actually use every day.
This represents years of trial and error, abandoned subscriptions, and the occasional bout of swearing at my screen. I’ve tested far more tools than I’ve kept, and the ones that survived did so because they deliver genuine value, and not because they have impressive marketing.
The Foundation: ChatGPT (The Swiss Army Knife)
ChatGPT serves as my all-rounder for quick chats, brainstorming, and general-purpose queries. I use it more for rapid-fire interactions than deep work. It handles everything adequately but doesn’t excel at anything specific, which is fine, because that’s precisely its role in my stack.
The feature I absolutely love is voice-to-text transcription. It lets me structure ideas on the move (walking to meetings, stuck in traffic, waiting for coffee) and have everything ready when I get to the office. The custom GPTs are also brilliant for specific use cases where I need consistent outputs without writing elaborate prompts every time.
The Workhorse: Claude (My Absolute Favourite)
Claude has become my primary tool for serious work. I pay for the Max account and find it the best value for money of all my AI subscriptions. It’s a genuine workhorse that handles everything from deep research to content creation tailored for different channels (social media, blog posts, client reports) all based on the learning structures I’ve built into my projects.
The coding capabilities are tremendous. I’ve tested many other tools, and some start well but then deliver subpar experiences. The debugging becomes mental. 🤦♂️
Claude’s extended thinking mode is brilliant for working through complex problems. It reasons through issues, finds new solutions, and actually accepts my corrections to make things work properly. Other tools? They nod along and then ignore everything you’ve said. 🙄
The Wild Card: Gemini (Brilliant but Unpredictable)
Gemini is something I’m still trying to decide whether I can fully trust. Not because of safety issues, but because the results are genuinely unpredictable. Sometimes it works beautifully. Sometimes it produces something absolutely bonkers that came out of nowhere. You’ll have three productive conversations about image editing, and then suddenly it shows you something that has no connection to anything you discussed.
The model can chunk vast amounts of data without much structure and give you useful findings. But it’s not linear, and I find it less predictable about what it will output next. That said, it’s absolutely amazing for image and video edits. Like a professional editor at the click of a button.
And here’s a trick I’ve discovered: create something beautiful in MidJourney, then ask Gemini to make an infographic with specific information inspired by that MidJourney example. Works like a charm. 😎
The Agentic Approach: Manus AI (When You Need to Think Big)
Manus AI is my go-to for agentic work. When I have a few ideas that I want to brainstorm and dump, and then see what emerges. It produces everything from short illuminating videos to stunning presentations with graphs and visual helpers. It reasons extensively, which sometimes feels like watching someone think out loud, but the outputs are worth the wait.
This aligns with what the research shows about agentic AI adoption. PwC’s 2025 survey found 79% of organisations have adopted AI agents to some extent, achieving 35% adoption in just two years—compared to eight years for traditional AI to reach 72%. Early data shows 66% of adopters report measurable productivity gains, with organisations projecting 171% average ROI from agentic deployments.
Presentations: Gamma (Making Sense of Complexity)
Gamma handles my presentations, course slides, and transforms complex data into simple visuals. I often use it after developing a structure in Manus or Claude. Feed it the thinking, and it creates stunning presentations on the go. For anyone who spends significant time preparing slides (guilty as charged), this tool pays for itself within a week.
Image Generation: MidJourney (Still the King)
MidJourney remains my primary image creator. 😍
I know Gemini can do outstanding work, and ChatGPT handles text-on-images brilliantly, but the consistency of brand style I get from MidJourney is unmatched. I can create images that people almost recognise as mine, even without seeing the brand. That’s the level of visual coherence it enables.
The small video animation feature (turning still images into subtle animations with one click) is sweeeet.
Super helpful for social content that needs a bit more life.
And the price? Absolutely ridiculous compared to other tools. In a good way. 👌
Video Editing: Veed.io (Scale Without the Headache)
Veed.io is my tool of choice for editing, creating, and even recording videos for social media and other channels.
It’s ridiculously easy to use and delivers powerful editing capabilities with AI assistance at the click of a button. When you know what you’re doing, you can produce multiple outputs in minutes.
That’s the kind of scale that actually matters when you’re running multiple brands.
AI Avatars: HeyGen (Still a Work in Progress)
HeyGen does what it promises (AI-generated video avatars) but requires you to understand its quirks.
The script is completely conditioned by what the tool might do when rendering the persona’s avatars.
You need to write around its glitches, which means knowing the small rendering issues in advance. Once you understand this, it works. Before that, expect frustration.
Automation: Make/n8n (Where the Real Magic Happens)
For AI Agents, I prefer deterministic structures with an agent’s ‘brain’ in between. I use specific AI Agents through automation tools like Make or n8n for campaign improvements, crunching financial numbers, preparing reports—anything I need on a regular basis with consistent outputs.
Nothing provides results like a fixed workflow for repetitive tasks.
The process is speedy, especially when you nurture those agents with RAG processes that keep adding knowledge to their reasoning without manual labour. This is where you stop playing with AI and start operationalising it.
Native AI Tools: The Hidden Gems
Here’s something people overlook: sometimes, instead of exporting data to ChatGPT or Claude, using the native AI tools within your existing platforms can be a shortcut to actually helpful outcomes.
I use Airtable‘s AI fields on my databases to keep information automatically updated, like searching for changes in enterprise fiscal information, tracking if companies have raised new funding rounds, and so on. ClickUp has AI helpers to keep projects and tasks in sync. Even Google Sheets has Gemini power for automatically update analysis over numbers. These integrations often work better than manual export/import cycles because they’re designed for their specific context.
Now, Hear Me Out
Let me share some things the AI evangelists won’t tell you, because admitting these truths would ruin their LinkedIn engagement rates.
Most ‘AI transformations’ are actually process fixes with AI sprinkled on top
BCG’s research confirms what I’ve seen in the field: the organisations getting real value from AI are the ones who had their processes sorted before they added AI. Manual optimisation must precede automation. If your process is broken, AI will just help you break it faster and at scale.
Congratulations 👏 👏 👏 👏, you’ve automated your dysfunction.
Simple workflows beat complex ones every single time
I see people building elaborate 47-step automation workflows with branching logic that would make a NASA engineer weep.
These workflows break constantly, are impossible to debug, and nobody remembers how they work six months later. The boring, simple workflow that does one thing well? That’s the one still running reliably a year from now.
The security situation is genuinely concerning
BigID’s 2025 research reveals significant gaps: 69% cite AI-powered data leaks as their top security concern, yet 47% have no AI-specific security controls. 64% of organisations lack full visibility into AI risks, and 55% are unprepared for AI regulatory compliance.
Shadow AI has become pervasive, with 29% of workers using AI who haven’t disclosed this to management, and 45.4% of sensitive AI interactions occur through personal accounts, bypassing corporate controls. 😬
Most businesses either over-restrict AI usage (killing productivity) or have no safety measures at all (inviting disaster). Finding the sensible middle ground is harder than it sounds.
Trust remains the limiting factor for agentic AI
Here’s a sobering statistic: HBR’s July 2025 survey found only 6% of companies fully trust AI agents to autonomously run core business processes. Most (43%) trust agents only with limited or routine operational tasks, while 39% restrict agents to supervised use cases.
Just 20% say their technology infrastructure is fully ready for agentic AI, and only 12% have risk and governance controls in place.
The hype around AI agents replacing entire departments is precisely that—hype.
We’re years away from genuine autonomous operation, and anyone telling you otherwise is either selling something or hasn’t actually tried to implement it at scale.
What’s Coming: Consolidation and the Death of Tool Sprawl
Gartner predicts 75% fewer generative AI vendors by 2029 through acquisitions and market failures. The current landscape (where the average enterprise marketing stack alone contains 90+ tools) is unsustainable.
Major productivity suites are embedding AI natively (Microsoft Copilot, Google Workspace), creating platform gravity that threatens standalone AI tools.
What does this mean for your tool selection today? Be thoughtful about vendor lock-in.
The specific tools matter less than your ability to adapt when the landscape shifts. Build processes that can survive tool changes. Document your workflows so you’re not starting from scratch when your favourite tool gets acquired or goes bankrupt.
Enterprise AI spending reached $37 billion in 2025 (up 3.2x from $11.5 billion the prior year), making it the fastest-scaling software category in history. Anthropic now holds 40% of the enterprise share, OpenAI 27%, and Google 21%. The race isn’t over, but the contours are becoming clearer.
The Bottom Line
We live in a strange and bright new world. The tools available today would have seemed like science fiction five years ago. But having access to powerful tools and knowing how to use them productively are very different things.
The professionals who will extract genuine value from AI are those who understand it as an operational challenge requiring organisational capability building, not a technology procurement problem to be solved by selecting the right vendor.
They’re the ones who fix their processes before automating them, choose purchased solutions over custom builds (67% success rate versus 22%, remember), focus 70% of their resources on people and processes, and start with governance frameworks before scaling.
My stack works for me because I’ve spent years figuring out what actually moves the needle for email marketing, AI automation, and B2B performance. Your stack will look different because your needs are different. The tools I’ve mentioned are far from magic. They’re instruments that become useful when wielded with skill and intention.
Start small. Automate the boring stuff first.
Get the quick wins before attempting the moonshot projects. And for the love of all that is holy, document what you’re doing so future-you doesn’t curse present-you when something inevitably breaks.
We can be super productive if we keep evolving and using the tools at our disposal. But productivity without purpose is just busy work with a better interface. Know what you’re trying to achieve, choose the tools that help you get there, and ignore the noise.
That’s been my approach for thirty years. It still works. 😎
Leave a Reply