From Web to Mobile to AI: The Design Lessons Everyone Keeps Forgetting

Designing Experiences at client workshops during 3 tech waves; Web, Mobile and AI

The headlines are clear: “90% of AI projects fail.” The MIT report confirms what many already sense — too many AI initiatives never make it past pilot stage.

But here’s the truth: AI doesn’t fail. Adoption fails!

Just like in the early days of the web and mobile waves, the same mistakes are being repeated: shiny pilots, weak integration, poor UX, and no real business alignment.

Let’s break it down.

Front-Office Agents: Visible, Popular… and Brittle

In my recent business tutorial on AI strategy and creating real P&L impact, any organization working towards an agentic AI state, the strategic choices are around two types of AI agents; front-office AI Agents or back-office AI agents.

Customer-facing, Front-end AI agents — bots for marketing, sales, or service — attract investment because they are visible. Boards love them, press releases love them, but customers? Not yet!

Why they fail:

• Integration gaps: often not tied into CRM or transaction systems.

• No memory: every customer query is treated in isolation.

• FAQ mindset: designed as glorified FAQs rather than workflow enablers.

• UX disappointment: poor conversational design drives customers back to human agents.

Result: expensive experiments with little P&L impact.

Back-Office Agents: Hidden, Valuable… and Stuck

Back-office AI agents — automating HR, finance, IT, or procurement — can be ROI heroes too. They reduce outsourcing costs, cut repetitive tasks, and create measurable efficiency.

Yet they fail too, for similar and different reasons:

• Legacy integration: old ERP/HRIS systems make seamless automation hard.

• Context gaps: agents can’t connect cases, invoices, or approvals.

• Lack of mature IT strategy; focus should be on data, model (including machine learning and training), infrastructure and AI governance.

• UX adoption: employees ignore tools that complicate their workflow.

• Governance fears: compliance and security concerns slow down adoption.

• Investment bias: most budgets flow to front-office “shiny” projects.

Result: hyped IT strategy and underfunded pilots that can never scale.

We analysed, based on the MIT report and other similar sources, the factors of recent AI failure. An overview is in the below table.

Factor	Front-Office Agents (Bots, CX, Marketing)	Back-Office Agents (HR, Finance, Ops)
Integration	Poor integration with CRM/ERP; often standalone pilots	Legacy systems, fragmented APIs, compliance-heavy integrations
Context & Memory	No memory of previous customer interactions → “dumb bots”	No workflow memory (cases, invoices, approvals) → frustration
UX & Adoption	Customers abandon bots quickly if UX fails	Employees ignore tools that add friction instead of reducing it
Strategic Alignment	Chosen for visibility, hype, and topline appeal	Underfunded despite stronger ROI potential
Governance & Security	Lower barrier, but brand risk if poor UX	Higher barrier (sensitive HR/finance data, compliance)
Investment Priority	Overfunded → but low sustainable impact	Underfunded → despite clear cost savings
Root Failure Pattern	Design flaws + customer adoption bottleneck	Integration complexity + governance hurdles

The Shared DNA of Failure

Whether front-office or back-office, the failure factors rhyme:

• Lack of integration

• No context or memory

• Weak UX and adoption

• No alignment with business outcomes

Different symptoms, same disease: bad design choices.

The Forgotten Design Lessons

We’ve been here before. In the web wave, clunky sites with no usability died fast. In the mobile wave, poorly integrated apps were deleted after one use.

The same is happening with AI. Without strategic design, your AI agent is just another pilot that won’t survive.

The key design lessons:

1. Workflow-first, not tech-first → start with the process, not the model.

2. Integration is everything → APIs are the lifeline of AI agents.

3. Memory matters → context turns a tool into an assistant.

4. UX drives adoption → if customers or employees hate it, it fails.

5. Measure business outcomes → design for ROI, not just engagement.

We have compiled another business tutorial outlining the fundamentals, 5 crucial design steps and conversational design frameworks to design AI agents, not just building.

With two large dependencies;

1.your IT strategy. Get the fundamentals right for your newly designed workflows or customer journeys.

2. Human-centered Design. Adoption of AI, including goes by the 10/20/70 rule. 10% is technology-related. 20% about systems and processes, 70 % is about human adoption. This is about Culture, Change (with limited resistance) and Customer Experience.

From Pilots to Agentic Enterprises

AI agents can transform businesses — but only if designed with strategy and discipline. At Appsolute Value, we capture these lessons in our Agentic Redesign Guide for AI Agents and use frameworks like the AI Business Value Matrix to help organizations make the right choices.

Because the next wave of winners won’t be the ones who “tried" the hype. They’ll be the ones who designed AI like they once designed great apps — with strategy, integration, and UX at the core.

👉 To get our Strategic Design Guides for AI Agents drop a DM at Michel van den Berg at Linkedin and explore the design frameworks that help you build AI agents — not experiments.