Where Will AI Be in 2 Years? What Business Owners Need to Know

Right now, AI agents — systems that don't just respond to prompts but autonomously plan, execute, and iterate on multi-step tasks — are in early adoption. Some businesses are using them. Most aren't. By 2028, that will have flipped.

Where will AI be in 2 years in terms of agentic capability? Agents will be handling entire workflows end-to-end. Not just “draft this email” but “read this inbound enquiry, research the prospect, qualify them against our criteria, draft a personalised response, schedule a follow-up, and log everything in the CRM.” That entire chain — currently requiring a human at every step — will run autonomously with human oversight only for edge cases.

The businesses that have already built their first automated workflows will be able to layer agentic capabilities on top of existing infrastructure. The ones starting from scratch will be building foundations while their competitors are already running at full speed. That's the compounding advantage of starting now — every system you build today becomes the base layer for something more powerful tomorrow.

Today, multimodal AI — models that process text, images, audio, and video together — is capable but still somewhat specialised. By 2028, it will be standard. Every major model will handle all modalities natively, and the cost of doing so will be a fraction of what it is today.

What does this mean practically? Processes that currently can't be automated because they involve reading handwritten notes, interpreting photos, analysing recorded calls, or processing video content will become routine automation targets. A tradesperson's job site photos can be automatically assessed and turned into scoped quotes. A recorded client call can be transcribed, summarised, action-itemised, and fed into your project management system — all without a human touching it.

Real-time processing is the other shift. AI that responds in milliseconds rather than seconds changes what's viable for customer-facing applications. Voice AI that handles inbound phone calls with natural conversation, live chat systems that feel indistinguishable from a human agent, real-time document analysis during client meetings — these are already emerging and will be mature by 2028. If you're still thinking about where will AI be in 2 years, the answer for real-time applications is: everywhere your customers interact with you.

AI API pricing has already dropped 10–50x over the past two years. The trajectory will continue. By 2028, what currently costs $100 a month in AI compute may cost $5. This isn't just a nice bonus — it fundamentally changes which processes are worth automating.

Today, there are processes in your business where the AI cost might not yet justify the automation — maybe the volume is low, or the task is moderately complex. In two years, the economics will flip on dozens of those edge cases. Processes that aren't worth automating today at $0.10 per transaction become obvious wins at $0.005 per transaction.

The businesses that have already built their automation infrastructure — their workflow orchestration, their data pipelines, their integration layers — will be able to expand their automation surface almost effortlessly as costs drop. The businesses that haven't built anything yet will need to start from the ground up, missing the window where falling costs create the easiest ROI opportunities.

Asking where will AI be in 2 years is really asking: what will my competitive landscape look like? And the answer is sobering. By 2028, the businesses that have embraced AI automation will be operating with significantly lower costs, faster response times, fewer errors, and better customer experiences than those that haven't. That's not speculation — it's already happening today, and the gap will only grow.

Operationally, this means several things. Your competitors who start automating now will have two years of compounding advantage by 2028. Two years of lower labour costs. Two years of faster turnaround times. Two years of data feeding into systems that get smarter with every transaction. Two years of their teams focusing on high-value work while AI handles the repetitive grind.

The compounding effect is the part most people miss. AI automation isn't a linear improvement — it's exponential. Each automated process gives you data, infrastructure, and operational knowledge that makes the next automation faster, cheaper, and more impactful. A business that automates its first process today and builds steadily over two years will have an automation capability that a late starter simply cannot replicate quickly.

There's a common belief that waiting is the safe option. “Let the technology mature. Let others go first. We'll adopt it when it's proven.” That logic made sense with previous technology shifts. It does not apply here.

With AI, waiting has a direct and quantifiable cost. Every month you delay automating a process that could be automated today is a month of unnecessary labour costs, a month of avoidable errors, and a month where your competitors are pulling further ahead. That's not a theoretical risk — it's money leaving your business.

The other risk is talent. As AI becomes standard business infrastructure, the people who know how to build, maintain, and evolve AI systems will be in high demand. Agencies, consultants, and internal hires with genuine AI automation experience are already getting harder to find. In two years, the supply-demand gap will be even wider. Businesses that build relationships with capable AI partners now will have priority access to expertise that latecomers will struggle to find.

The technology is already proven. The ROI is already clear. The question is no longer “does this work?” — it's “how much ground am I losing by not acting?”

Understanding where AI will be in 2 years is useful. Positioning your business to benefit from it is what matters. Here's what that looks like in practice.

Start with one high-value automation now. Not a pilot. Not a research project. Pick your most expensive manual process and automate it properly. Build it on infrastructure you own — tools like n8n and Supabase that give you full control and flexibility. Get something live, measure the results, and use the win to build momentum for the next one.

Build for flexibility, not just for today. The AI models you use today will be replaced by better, cheaper ones within months. Your architecture needs to accommodate that. Design systems where the AI layer is swappable — so when a new model drops that's faster or more accurate, you switch a configuration rather than rebuild the system. This is exactly what a solid AI integration strategy delivers.

Develop internal AI literacy. Your team doesn't need to become AI engineers. But they need to understand what AI can do, how to work alongside automated systems, and how to spot opportunities for new automation in their daily workflows. A team that actively feeds ideas into your automation pipeline is worth more than any single AI system.

Partner with someone who lives in this space. AI is evolving too fast for anyone outside the industry to track effectively. You need a relationship with an agency or partner who is actively monitoring what's new, what's practical, and what's overhyped — and who proactively brings recommendations to you. Quarterly reviews, system upgrades, new capability assessments. The businesses that stay ahead are the ones with someone watching the horizon on their behalf. We wrote about this at length in our guide to future-proofing your business with AI.

Think in terms of compounding, not one-off wins. Every automation you build today is infrastructure for tomorrow. The data it generates trains better systems. The integrations it creates become the backbone for more advanced workflows. The operational knowledge your team gains compounds with every project. Two years of steady building creates a capability that no amount of money can buy overnight.