Cutting through the AI noise: six realities for 2026

Last week we hosted a webinar with our founder Jack Rimmer and technical director Frederick Grover, with the goal of cutting through the hype and sharing a pragmatic view on where AI, automation and low-code are genuinely driving value this year (as well as where caution is warranted).

Here are the six themes our speakers said leaders should focus on in 2026:

1. Signal over noise: AI is here to stay - make it relevant to your business

Usage is exploding. Tools like ChatGPT now handle billions of prompts daily and have become part of the modern workday. The imperative is focus. Instead of headline-grabbing experiments, target AI where it can move a real metric for your organisation - cost, speed, risk, revenue, CSAT etc.

2. RPA isn’t dead - it’s a “useful past” and a reliable fallback

RPA exists to bridge old systems, missing APIs and rigid processes with dependable, repeatable execution. What’s new is that computer-vision agents and LLM-powered UI automation are creating competition at the edges. In reality - when you need deterministic, governed workflows at scale, RPA still wins on cost, predictability and auditability. As Jack quoted Microsoft’s Ryan Cunningham, “We didn’t buy an RPA tool because we thought it was the future, but we recognised it’s a useful past.”

3. Agentic AI: do you need it (yet)?

The hype: “Define a goal, set guardrails, let the agent figure it out.” But the business truth is that most enterprises don’t hire a new customer service rep, hand them a target, and walk away. We provide procedures, decision points, and escalation paths. The middle path would be automated AI workflows. Combine structure (well-defined steps, audit trails) with generative components where judgement is needed. Tune freedom based on context - regulated teams need tighter controls than, say, creative marketing squads.

4. ROI over buzzwords: start with the problem, not the tool

Robiquity’s rule of thumb is that if we can’t show a material 3:1 return on a programme, we’re on thin ice. You should avoid ‘hammer-looking-for-a-nail’ programmes driven purely by vendor funding or hype, but instead revisit backlogs. Many previously marginal use cases now clear the bar with LLMs in the mix - delivered faster and to a higher standard than a few years ago.

5. Over-reliance is a real risk: move from “human-in-the-loop” to “human-in-command”

Copy-paste coding from AI without understanding leads to spaghetti code and technical debt are a real behavioural risk. Search results with AI answers can accelerate shallow adoption and errors. Over-trusting outputs just because they come from “the system” Is another risk. The UK Post Office Horizon scandal is a cautionary tale about deference to software. Look to change the mindset. A human clicking “approve” in a queue is not control. Assign human-in-command roles with accountability for outcomes, not just checkpoints. Rotate responsibilities to avoid rubber-stamping and complacency; keep fresh eyes on critical flows.

6. Governance and orchestration: clarify ownership, build guardrails, modernise the “glue”

Many organisations default AI risk to data protection leads. That’s necessary (data sovereignty, lawful basis, model usage), but not sufficient. Someone must own operational risk from AI-assisted decision-making and over-dependence. Regulators are reactive. Don’t wait for perfect guidance. Define internal policies now - acceptable use, logging, model selection, evaluation, red-teaming, and incident response.

Go beyond single-stack thinking. Heavy RPA estates can struggle with modern orchestration. Move towards best-fit tooling and intelligent orchestration across AI agents, cloud flows, APIs and RPA. Aim for responsive, event-driven architectures with strong observability.

Closing takeaways from Jack and Fred:

• Problem-first beats tool-first. Don’t let vendor hype set your priorities.
• Reassess old backlogs with a fresh LLM lens - many now clear the ROI bar.
• Don’t throw away proven ways of working. Bring AI into structured workflows with clear decision and escalation points.
• RPA isn’t dead; it’s part of a balanced, governed automation strategy. Ask us again next year - but for now, it still matters.

If you’d like a deeper dive on where AI, automation and low-code can drive a 3:1 return in your environment - and how to govern it safely - we'd love to har from you.