Virtually every week I get asked by another client whether they should go all-in on Agentforce.
No, they shouldn’t.
It’s a slightly unnerving answer from an AI consultant in the Salesforce world… but I firmly believe that locking down to a single AI provider would be a suicidal decision for any business.
The reality is that AI innovation moves so fast that any strategy locked into a single tool will be outdated before you can say the word ‘Agentblazer’. AI has been moving on a timescale that makes most business planning frameworks look glacial.
Last year’s breakthrough becomes this year’s open-source commodity, often within weeks.
As BCG bluntly put it:
“What feels advanced today will be table stakes by 2030 – if not before”.
Personally, I think they’re even underestimating the pace of change.
In other words, by the time you’ve finished debating the ROI case with 5 levels of management, the thing you’re debating is already outdated.
I’ve watched SaaS vendors proudly showcase a shiny new AI feature only to discover it’s been leapfrogged before the rollout is complete.
Ultimately, the shelf life of an AI edge is shrinking to a year or less.
Agility, not certainty
So what does this mean for strategy?
It means you don’t build AI into your business like you’re buying heavy machinery that needs a 10-year payback. You design for agility and adaptability.
Treat models and vendors as interchangeable parts, not foundations. That creates a focus on shorter development sprints, modular platforms, and a portfolio approach to AI investments.

In other words, don’t marry your vendor; date them casually, and always keep your options open. Speed beats perfection here.
One practical example of this strategy in action is that you want to be able to swap in the next model breakthrough without breaking production, a concept known as ‘hot-swapping’.
This isn’t a completely new concept, an agile IT strategy became popular (and was enabled) by the shift to the cloud but many businesses have gotten away without really embedding any agility. I don’t believe AI will be so forgiving.
Looking through the four lenses of AI
When I talk about where AI is heading, I find it useful to look at four of the most visible lenses: video, image, content, and predictive.
Each one gives an insight into how businesses will need to reinvent themselves.
Video.
Generative video is already blurring the line between filming and rendering. Platforms like Synthesia can produce realistic avatars from nothing but a script, while D-ID lets you translate live speech into another language in real time.
For training, marketing, and corporate comms, this has the potential to mean global reach at a fraction of the cost.
I’ve had plenty of clients half-jokingly suggest building a “digital David” to handle my meetings, workshops, and training sessions. I told them the real one still comes with the occasional (good) bad joke and sarcasm, something AI can’t quite replicate yet.
Go ahead and create an AI avatar on HeyGen then tell me that it ‘won’t happen’.
Image
AI image generation has gone from experimental novelty to commercial-grade production. Platforms like Midjourney, Ideogram, and Runway can now produce near-photorealistic visuals tailored to a brand’s exact brief. For design and marketing teams, that means product imagery, campaign assets, and creative concepts can be generated in hours instead of weeks.
Retailers such as Zalando and IKEA are already experimenting with AI imagery for catalogue production, enabling rapid localisation and seasonal refreshes without the need for reshoots. Automotive brands are also starting to create entire photo libraries of concept vehicles without a single prototype leaving the factory.
Content
Generative text has evolved from assisting writers to operating as a full content engine. Tools such as Writer, Jasper, and Notion AI integrate brand tone, compliance rules, and workflow approvals to generate material that’s both creative and consistent.
Large enterprises are deploying these systems to draft press releases, training modules, and localised marketing collateral. Unilever and Estee Lauder have both piloted AI content pipelines that produce region-specific communications aligned to global standards.
Most brands aren’t talking publicly about this to try and preserve some competitive advantage because the barrier to entry is so low.
One example I absolutely love is Spotify’s use of LLMs within their annotation process, where humans are still in the loop but AI allows them to operate at a previously impossible scale.

Predictive
Predictive AI is transforming decision-making from hindsight to foresight. Advanced platforms like C3.ai, DataRobot, and Amazon Forecast are being used to anticipate customer behaviour, optimise logistics, and manage demand variability.
Stop reacting to asset failures. Start predicting them.
C3 AI Asset Performance Suite uses enterprise AI to maximize uptime and optimize maintenance.
See it in action:
— #C3 AI (#@C3_AI)
4:50 PM • Jul 28, 2025
In automotive, predictive maintenance systems such as Uptake and Pitstop are analysing sensor data to detect issues before failures occur. In retail and manufacturing, companies like Snowflake and Databricks are building real-time prediction pipelines that integrate live CRM, ERP, and IoT data streams to forecast outcomes as they happen.
Instead of waiting for monthly reports to tell you what happened, it’s possible to have an AI that’s already adjusting inventory, rerouting shipments, and flagging the customer who's about to churn.
The “startup rebuild” exercise
Here’s a thought experiment I use with leadership teams: imagine your company collapsed last night, and today you’re rebuilding it as a startup using AI from day one.
What stays?
What goes?
Creating this environment gives everyone some space to re-focus on the actual value their business is creating, and flags many of the distractions that could now be (semi-) automated.
A logistics firm might suddenly realise that half its headcount is tied up in scheduling and routing that a modern competitor would automate immediately.
A manufacturer could eliminate entire layers of QA and maintenance.
A consulting firm would finally admit it doesn’t need staff trawling through PDFs when AI can do it faster and more accurately.
The uncomfortable truth these examples expose is that much of what we defend as "how we do business" is really just "how we used to have to do business".
The point isn't to fantasise about startups (brutally hard in their own right), it's to help identify how much of your current structure exists only because better options didn't exist when you built it.
The most common counter-argument I get at this point normally revolves around the fact AI doesn’t take the right actions 100% of the time. My response is simply that neither do humans, we’re just emotionally uncomfortable handing over decisions to a non-human entity.
Instead, we should be objective in our assessment of humans vs AI. Something I’ll try and dive deeper into with another article. This one’s getting a bit long.
Implications for leaders
So, what does this actually mean in practice? I think there’s three big takeaways.
First, shift your horizon to 12–18 months. Forget five-year AI plans. The ground will shift underneath you long before then. Instead, run scenarios: what does your business look like when 80% of admin is automated? How does your industry change when predictive AI gets prescriptive?
Second, invest in adaptability, not perfection. Don’t overcommit to closed solutions when open models can leapfrog them in weeks. Build modular IT and data systems so you can swap components in and out. Value optionality over lock-in.
Third, build a culture of perpetual learning. That means upskilling teams, encouraging experiments, and accepting that failure is part of the process. The companies that thrive won’t be the ones who “got AI right” once. They’ll be the ones constantly reinventing their playbook.

Living in “AI years”
It’s important to remain aware that today's AI is the worst it will ever be.
Tomorrow's models will be more capable, more efficient, and more integrated into our work. Stability isn't just an illusion, it's a competitive disadvantage.
The real risk isn’t spending on the wrong tool, it’s designing your business for a competitive market that no longer exists.
Personally, I’ve come to find it freeing. If nothing is permanent, then we’re all startups again. The better question isn’t “What’s our AI strategy today?” It’s “If we had to rebuild tomorrow, what would we do differently?”
That’s the question I keep asking myself and the one I think everybody should spend a few days sitting with and mulling over. My belief is that your success over the next 10 years will likely correlate with whether you find that thinking uncomfortable or inspiring.
I’m inspired, are you?
I’m David Bruce, and this is Optimistic Intelligence.
I write about AI from the perspective of someone who’s been consulting for some of Australia’s biggest businesses over the last 5 years (and another 5 at one the world’s biggest FMCG firms)… including the good, the bad, and the occasionally ridiculous.
The goal’s simple: cut through the hype and talk honestly about how AI is really showing up in work and consulting. This is a space for me to articulate the speed at which the world is changing around us, trying to wrangle the chaos and ground it in reality.
I aim to write one article a week, sometimes more, sometimes less.
If that sounds useful (or at least more interesting than another daily AI news summary email) hit subscribe and join me.