LuminateCX Blog

"

Is your AI strategy delivering results, or just running up the bill?

OpenAI has announced it will deprecate GPT-4.5 in July. It costs 30 times more than GPT-4o, yet offers only marginal gains. This isn't just a pricing change. It's a signal.

Many organisations are still deploying AI like it's 2022, one large model, stretched across every use case. That approach isn't scalable or cost-efficient. It's a liability.

Today, there's increasing access to fit-for-purpose models from open-source options like Mixtral and LLaMA to enterprise tools like Claude, Gemini and GPT-4o. Each brings different strengths depending on the task, domain or cost to value.

The companies getting this right aren't choosing "the best model." They're orchestrating across many. They route simple prompts to efficient models and reserve premium power for high-value use cases.

This article breaks down the real cost of single-model thinking and what it takes to build a smarter, leaner AI strategy in 2025.

The Era of One-Model AI Is Over

Not long ago, deploying AI meant choosing a model and pushing it across as many use cases as possible. That strategy might have been passable in 2022, but with what we have available now, it's not just an outdated approach, it's an expensive one at that.

We've moved from only having access to a handful of generalist models to many purpose built models and although this could be seen as noise, it's more of a lever waiting to be orchestrated. Even within just the big players  we see that OpenAI offers GPT-4o, o4-mini and 5+ other models on their platform. Google provides Gemini 2.5 in both "Pro" and "Flash" editions as well as 10+ other variants. Anthropic's Claude 3.7 Sonnet introduces hybrid reasoning variants and flexible pricing and access to 3+ other variants of its models. 

What matters isn't just the name or brand. It's that the cost-performance delta between these models is now massive.

• GPT-4.5: $75 per million input tokens, $150 for output, with only 7% performance improvement over GPT-4o (Ojha, 2025; TechCrunch, 2025)
• GPT-4o: $2.50 input / $10 output per million
• Claude 3.7 Sonnet: $3 input / $15 output, and it handles multi-step reasoning (Anthropic, 2025)
• Gemini 2.5 Pro: $2.50 input / $15 output per million
• Gemini 2.5 Flash: $0.15 input / $0.60 output for fast-response tasks 

Behind the Curtain: Smarter Models, Not Just Bigger Ones

The evolution of AI models in 2025 is marked by a strategic shift towards Mixture of Experts (MoE) architectures, with a focus on efficiency and specialisation over brute-force scale.

Meta's Llama 4 series, released in April 2025, exemplifies this approach. Models like Llama 4 Scout and Maverick use MoE designs that activate only a fraction of their total expert pathways per task. For example, Llama 4 Maverick includes 128 experts, but uses just 17 billion active parameters per inference, a configuration that significantly reduces compute costs while retaining high performance (Dataconomy, 2025).

But perhaps more importantly, enterprises are now beginning to design and fine-tune bespoke MoE models for specific domains: customer service, finance, compliance triage. These tailored models don't aim to outperform general-purpose giants like GPT-4. Instead, they deliver 90% of the value at a fraction of the cost, while aligning tightly with business objectives.

This isn't just a technical upgrade. It's a strategy shift away from "best model overall" toward "best model for this job, at this price."

As model orchestration matures, modularity becomes a competitive lever. Companies embracing smaller, task-specific MoEs are gaining precision, control, and commercial agility.

What Orchestration Looks Like in Practice

Some companies have already made the shift.

A major auto manufacturer uses Microsoft's lightweight Phi model to handle the bulk of incoming customer queries. Only the edge cases get routed to GPT-4o. The orchestration reduced their AI spend by 80 percent, while improving response quality.

Sage, the accounting software company, fine-tuned Mistral for their domain, using it as a triage layer before escalating to a larger model. It improved resolution accuracy and delivered faster responses.

Vendors are starting to catch up:

• Azure AI Studio now offers over 1,800 models with routing capabilities
• AWS Bedrock provides a unified interface for multiple foundation models
• Google Vertex AI hosts more than 200 models in its Model Garden

Startups like Martian are building routing layers that dynamically switch models based on task complexity. Customers report 70 to 90 percent reductions in model costs with no drop in output quality (TechCrunch, 2025).

Two Strategies, Same Use case, Radically Different Outcomes.

Let's look at how this plays out in practice. Below, two companies deploy models across similar use cases, but only one does it with orchestration in mind.

*Pricing approximate & in USD. Source: OpenRouter.ai

The only difference is that Company B didn't just choose models, they built a model strategy.

What Smart AI Strategy Looks Like

The orchestration approach isn't about jumping on the next new model. It's about building a framework that supports:

• Task audits – Start with what each part of your business needs
• Cost-to-value mapping – Not every prompt deserves frontier-level performance
• Routing logic – Route based on complexity and risk, not convenience
• Ongoing optimisation – Measure performance/cost by model and task type and adjust routing rules accordingly.

This is what separates maturity from experimentation. Without it, you're not building AI systems, you're running expensive experiments at scale.

Time to Unpack What's Really Happening

Companies still relying on single-model AI deployments will be running tech strategies from a previous era. It's no longer enough to have "an AI model." You need a way to manage, optimise, and evolve your AI footprint in line with business goals.

"

The winners in this economy won't have the best model. They'll have the clearest orchestration strategy.

References 

Anthropic. (2025). Claude 3.7 Sonnet and Claude Code. https://www.anthropic.com/news/claude-3-7-sonnet 

Ars Technica. (2025). "It's a lemon": OpenAI's largest AI model ever arrives to mixed reviews. https://arstechnica.com/ai/2025/02/its-a-lemon-openais-largest-ai-model-ever-arrives-to-mixed-reviews/ 

Meta / Llama 4:
Dataconomy. (2025, April 7). Meta launches new Llama 4 AI models: Scout, Maverick now available in apps. Retrieved from https://dataconomy.com/2025/04/07/meta-launches-new-llama-4-ai-models-scout-maverick-now-available-in-apps/?utm_source=chatgpt.com 

Mistral / Mixtral MoE:
Artetxe, M., Rabe, M. N., Scao, T. L., Tunstall, L., de Masson d'Autume, G., & Elbayad, M. (2024). Mixtral of MoE: Sparse mixture of experts models with open weights (arXiv:2401.04088). arXiv. https://arxiv.org/abs/2401.04088 

Google Cloud. (2025). Vertex AI Pricing. https://cloud.google.com/vertex-ai/generative-ai/pricing 

Ojha, A. (2025). The great paradox: Why OpenAI's most expensive model GPT-4.5 falls short of expectations. https://medium.com/@ayushojha010/the-great-paradox-why-openais-most-expensive-model-gpt-4-5-falls-short-of-expectations-4c3c5035a692 

TechCrunch. (2025). OpenAI plans to phase out GPT-4.5. https://techcrunch.com/2025/04/14/openai-plans-to-wind-down-gpt-4-5-its-largest-ever-ai-model-in-its-api/ 

"

AI doesn't know your brand. Or your people. It can learn patterns, generate content, even sound convincing. But it has no memory of what your team went through last year. It doesn't understand your customers' frustrations. It doesn't carry your culture.

We're standing at a strange intersection. AI is everywhere, embedded in our workflows, our tools, even our job descriptions. Yet for many business leaders, it still feels distant. Shiny. Unsettling. Unproven.

In the rush to integrate AI, the narrative has been hijacked by tech vendors and productivity metrics. Platforms promise acceleration, automation, and insight at scale. But the question we're wrestling with at LuminateCX is more human:

As we prepare for Humanising AI Brisbane 2025 and mark one year of LuminateCX, I've been reflecting on what "humanising AI" really means. Not just as a concept, but as a strategy. A leadership responsibility. A design choice.

When AI Strategies Fail, It's Rarely About the Tech

Most AI pilots don't stall because the tools weren't smart enough. They stall because the people weren't brought along.

We've seen this play out in real-time: leadership teams chasing AI initiatives without first answering why it matters to their people, how it fits their culture, or what problems it actually needs to solve.

"

An AI strategy does not start with just simply buying a piece of technology and implementing it employee-wide. The true strategy is to first look inwardly — to your people and your processes.

That's not just a philosophical stance. It's a pattern we've observed across dozens of organisations.

• Tech teams light up a GPT tool... but nobody uses it.
• Marketing automates content... but brand voice gets diluted.
• HR explores AI-powered onboarding... but employee trust erodes.

Too often, AI is introduced as a layer on top of the business — not within it.

And when that happens, three things suffer:

"

Simply turning on AI capability doesn't mean people will use it. Unless they see the value — how it solves a problem or benefits them — it's just another tool gathering dust.

So What Does Humanising AI Look Like?

It starts small  and it starts with people.

The term "humanising AI" gets thrown around a lot. But in practice, it's not about warm fuzzy branding or philosophical debates. It's about design. Leadership. Practical choices in how AI is introduced, structured, and scaled across the organisation.

"

My initial take of what humanising AI means is really, how do I incorporate this latest technology into helping me do what I find is important — for me, for my job, for the people and systems I work with.

At LuminateCX, we've found that the most successful AI adoption stories don't begin with automation targets, they begin with understanding friction. Where are people stuck? What drains their energy? What could help them feel more effective?

Humanising AI means looking at your workflows and asking:

• "Where can we remove complexity without losing connection?"
• "What micro-tasks are ripe for augmentation, but still need a human in the loop?"
• "How do we build trust in these systems through visible wins?"

"

Going into your organisation and asking 'what can we automate?' is not a recipe for success. It pushes back on your people.

Instead, the better question is:

This approach does more than drive productivity. It builds momentum. You're not just integrating tech,  you're creating buy-in, enabling experimentation, and embedding AI into your culture without breaking it.

"

Where there are small wins that are of value to automate — where the expert in the middle still plays a role — that's where AI adoption becomes trusted, useful, and human.

This Isn't About the Future. It's About Right Now

We've moved past prompts. AI isn't something we experiment with on the side anymore, it's operating behind the scenes, shaping decisions, streamlining processes, and driving change whether we plan for it or not.

"

Those systems will increasingly become almost like organisational AI. They won't just respond — they'll orchestrate. Multi-step tasks, decision-making flows, process automation... carried out under the hood, with more abstraction at the surface.

That abstraction means less visibility and more risk.

If you don't actively shape how AI behaves in your organisation, it will still operate… just without your values in the loop.

That's when things start to break.

Brand gets diluted. Trust erodes. People start checking out — or worse, they bypass the system entirely.

That's why humanising AI isn't a philosophical nice-to-have. It's a strategic necessity.

You don't need to have all the answers. But you do need a plan.

"

Right now, the opportunity is to bring all of your teams to a base level of competency on the use of AI. If you're not doing it, your competitors are.

So Where Do You Start?

Not with a platform. Not with a pilot.

Start with your people.

"

AI doesn't know your company. It doesn't understand your culture, or your customers. Your people do. There's so much richness in their experience — and that's what should shape how AI is used, not the other way around.

The Humanising AI Approach: A Practical Flow

At LuminateCX, we've seen that humanising AI isn't a one-off initiative. I t's a sequence of leadership choices made over time. Here's how we frame it:

You don't need a perfect strategy on Day One. But you do need to move and move in a way your people can believe in.

Let's not sugar-coat it.

If your AI pilot went well, there's a good chance it's headed straight for the scrapheap.

Gartner just made it official: by the end of 2025, 30% of generative AI projects that prove successful in pilot will be abandoned before reaching production. Not because the tech didn't work—but because the business wasn't ready.

The Pilot Trap: Why Early Success is Misleading

Successful AI pilots are designed to prove the art of the possible—not the rigour of the real.

They're built on:

• Handpicked, well-labelled datasets
• Controlled use cases
• Isolated environments with minimal risk
• Tight-knit project teams with deep support
• Zero integration with legacy platforms

But once the pilot ends, reality begins.

Suddenly you're facing:

• Disparate data quality across business units
• Conflicting priorities between IT, ops, and execs
• Integration complexity with platforms that weren't part of the pilot
• Governance, compliance, and risk questions that never got asked
• A board expecting results while your team scrambles for alignment

GenAI Pilot Purgatory: The New Enterprise Bottleneck

HFS Research coined the term "GenAI Pilot Purgatory" to describe this exact scenario:
Where innovation gets stuck between validation and value.

The technology works—but the business isn't built to scale it.

We've seen this play out in both public and private sectors. Pilots get board-level airtime, but when it's time to integrate with mission-critical systems, the gaps appear. AI initiatives are derailed not by model drift, but by organisational friction.

McKinsey summed it up well:

"

Business leaders face increasing pressure to generate ROI from their GenAI deployments.

And that pressure is real. Especially when every other line item in your budget is being challenged.

Scaling AI Requires More Than a Rollout Plan

Here's the common failure pattern we see:

It's the equivalent of building a show home before securing zoning approval.

From Pilot to Production: The Ignite Approach

At LuminateCX, we built Ignite to help organisations avoid exactly this scenario.

Here's how we do it:

What the Data (and Our Experience) Actually Tells Us

AI success isn't about picking the right tool—it's about building the right conditions for scale.

We've seen the pattern across sectors—whether it's a government agency piloting language models for service delivery, or a national insurer testing claims automation. The tools work in isolation. The frameworks often don't exist.

And while few organisations have mastered AI at scale, the red flags show up early:

• Governance gaps
• Shadow adoption without oversight
• Mismatched expectations between tech and business units
• No clear ROI model beyond the pilot phase

At LuminateCX,  our team have spent their careers guiding clients through complex digital transformations—across MarTech, data integration, and cloud migration. While GenAI is newer, the underlying transformation risks are familiar.

We know what it looks like when frameworks are missing—and we know how to build the ones that unlock momentum.

You've Got 8 Months

That's how long you have before Gartner's prediction becomes your reality.

Eight months to either:

• Join the 30% of organisations that never move beyond the pilot
• Or build a foundation that makes AI a core capability—not a proof of concept

Your Next Strategic Move

Before you double down on scaling your AI pilot, take a day to pressure-test your readiness.

Book a Spark Session—a structured, advisory-led engagement that identifies the key blockers, risks, and opportunities before you commit capital or credibility to a flawed rollout.

Let's ensure your pilot is the starting line and not the finish.

Introduction

When OpenAI released its AI in the Enterprise guide, it offered a compelling blueprint for AI adoption. The seven lessons—drawn from companies like Klarna, BBVA, and Morgan Stanley—are instructive and well-articulated.

But for most enterprises, especially in regulated industries or mid-market environments, getting from AI pilot to business-as-usual (BAU) requires more than inspiration. It calls for a structured path, organisational readiness, and a mindset shift. This article offers a strategic reflection on OpenAI's lessons, adding context, caution, and concrete steps. The goal: help enterprises move from experimentation to embedded capability.

1. Start with Evals — but Choose Metrics That Matter

OpenAI recommends: Start with systematic evaluation processes to measure how AI models perform against use cases.

Why it matters
Morgan Stanley's early success with generative AI stemmed from carefully structured model evaluations (OpenAI, 2024). By rigorously assessing summarisation, translation and relevance, they created confidence to scale.

LuminateCX Insight: Many enterprises over-index on model benchmarks without linking them to operational impact. A model might hit 95 percent accuracy and still fail to change outcomes. Without alignment between eval metrics and business KPIs, evals create false confidence.

2. Embed AI in Products — but Rethink the Journey Too

OpenAI recommends: Embed AI into your products to create smarter, more responsive customer experiences.

Why it matters
Indeed's use of AI to explain job matching improved engagement and conversion, driving a 20 percent increase in applications started (OpenAI, 2024).

LuminateCX Insight: Too often, enterprises layer AI features onto existing tools without rethinking the broader experience. This leads to confusion or degraded UX. The AI feature may be technically impressive but not functionally useful.

3. Start Early — but Only If You're Ready

OpenAI recommends: Invest and start early to gain compounding value.

Why it matters
Klarna's early investment led to an AI assistant now handling two-thirds of customer chats, delivering a projected $40 million in efficiency gains (OpenAI, 2024).

LuminateCX Insight: Some organisations start before governance, security, or alignment is in place. Samsung's data leak via ChatGPT use by engineers led to an internal ban and reputational risk (TechCrunch, 2023).

"

Starting early is not the same as starting recklessly.

4. Fine-Tune When Necessary — but Don't Default to It

OpenAI recommends: Customise and fine-tune models for specific use cases.

Why it matters
Lowe's improved product tagging accuracy by 20 percent through fine-tuning GPT models (OpenAI, 2024).

LuminateCX Insight: Fine-tuning increases operational complexity. It creates maintenance overhead, locks in assumptions, and limits flexibility when base models evolve. Fine-tuned models can also become outdated quickly and require constant retraining (OpenAI Community Forum, 2024).

5. Empower Experts — but Ensure Support Structures Exist

OpenAI recommends: Put AI in the hands of the people closest to the process.

Why it matters
BBVA enabled over 125,000 employees to create their own GPT-powered apps. This decentralised innovation accelerated adoption and surfaced unexpected value (OpenAI, 2024).

LuminateCX Insight: Giving people access to AI doesn't mean they will use it effectively. Many teams lack training, support, or incentives. AI tools without enablement lead to inconsistent adoption.

6. Unblock Developers — and All Other Stakeholders

OpenAI recommends: Automate software development processes to accelerate AI delivery.

Why it matters
Mercado Libre built a platform (Verdi) that allowed over 17,000 developers to accelerate AI-enabled applications, improving inventory, fraud detection, and customer experience (OpenAI, 2024).

LuminateCX Insight: Bottlenecks rarely exist in development alone. Security, legal, procurement, and compliance often delay deployment—especially in regulated industries.

7. Set Bold Automation Goals — but Stay Grounded

OpenAI recommends: Aim high with AI. Don't limit yourself to low-hanging fruit.

Why it matters
Ambitious automation can transform processes and free capacity. OpenAI automated internal support workflows, increasing efficiency (OpenAI, 2024).

LuminateCX Insight: Overpromising AI potential without operational discipline leads to failed projects and executive fatigue. Not all processes are ready for automation.

From Pilot to BAU: What It Really Takes

Moving AI into BAU doesn't just mean deploying it. It means:

• Clear ownership and accountability
• KPIs linked to operational or customer outcomes
• Change management and communication
• Ongoing model monitoring, retraining, and compliance review

LuminateCX Insight: AI in production must meet the same standards as any enterprise system: secure, supported, and sustainable.

LuminateCX BAU Readiness Checklist

Before scaling an AI use case, ask:

If the answer to any of the above is "no," the project may not be ready to move to BAU.

Conclusion

OpenAI's seven lessons are a valuable starting point. But AI adoption doesn't succeed through vision alone. It requires readiness, execution discipline, and clarity on what "good" looks like post-pilot.

"

For enterprise leaders, the challenge is not in launching AI pilots—it's in embedding AI in ways that are safe, valuable, and scalable.

For enterprise leaders, the challenge is not in launching AI pilots—it's in embedding AI in ways that are safe, valuable, and scalable. That's what separates AI as hype from AI as a lever for transformation.

References

• OpenAI. (2024). AI in the enterprise: Lessons from seven frontier companies. Retrieved from https://openai.com/business/
• Weill, P., Woerner, S., & Kiron, D. (2025). What's your company's AI maturity level? MIT Sloan Management Review. Retrieved from https://mitsloan.mit.edu/ideas-made-to-matter/whats-your-companys-ai-maturity-level
• TechCrunch. (2023, May 2). Samsung bans use of generative AI tools like ChatGPT after internal data leak. Retrieved from https://techcrunch.com/2023/05/02/samsung-bans-use-of-generative-ai-tools-like-chatgpt-after-april-internal-data-leak/
• OpenAI Community Forum. (2024). Fine-tuning myths and usage guidance. Retrieved from https://community.openai.com/t/fine-tuning-myths-openai-documentation/133608

The Real Shift: It’s Already Happened

Most teams are still chasing AI disruption. But the sharpest operators have already moved on because the real disruption? It’s behind us.

While the market obsessed over prompts, copilots, and generative UI tricks, something quieter but more consequential took root: a fundamental shift in where execution happens.

The execution layer has moved. It’s no longer buried in your dev backlog, your delivery pipeline, or integration stack. It now lives at the edge, where intent meets outcome, instantly.

This is the Zero-to-Solve era: a world where business users, strategists, and domain experts don’t just generate ideas, they deliver them. Not with dev tickets, but with AI-native tools that turn prompts into products, ideas into automation, and bottlenecks into velocity.

From Curiosity to Capability

Remember when GPT and Claude first hit the mainstream?

We treated them like clever assistants, smarter search engines, better writers, AI sidekicks to help tidy content or clean up code.

But the true inflection point wasn’t when they started speaking back. It was when they started executing unprompted, unassisted, and with intent.

Take Claude’s Artifacts feature. What looked like a UI enhancement was something far deeper: the emergence of AI as a systems thinker. The model wasn’t just answering, it was coding, styling, and structuring the output inside an interactive canvas without being explicitly told to (Anthropic, 2024).

That’s when the shift happened.

We stopped prompting, and started prototyping. In real time. With real outcomes.

And just like that, we crossed the line from disruption to democratisation.

Zero-to-Solve: Execution at the Speed of Intent

Zero-to-Solve isn’t a feature. It’s a new operating model where execution moves at the speed of thought.

No more waiting on prioritisation, pipelines, or tickets. If you can describe it, the system can deploy it.

One user built and launched a fully functional note-taking app on Bolt.new in under two minutes—starting from a blank canvas, using nothing but plain language prompts (The Prompt Warrior, 2025).

This isn’t code generation. It’s product delivery.

It’s what happens when infrastructure fades into the background and intent becomes the API.

We call this vibe coding. A new paradigm where the builder defines the outcome, and the system handles the logic, syntax, and deployment path (Garg, 2025).

Describe what you want. The stack responds. Zero-to-Solve doesn’t just accelerate workflows. It rewrites them.

The Rise of the Citizen Developer

We used to say, “If you can dream it, you can build it.” But that used to mean: funding, engineers, velocity planning, and three months of backlog wrangling.

Now, you describe it. And it ships.

AI-native platforms have collapsed the gap between idea and execution. Non-technical users, product managers, marketers, operations leads, are turning plain language into production-ready tools, workflows, and systems. No code required. No approvals queued.

The new builder stack includes platforms like:

• Windsurf: An agentic IDE that applies multi-file edits using Flows and Cascade (Codeium, 2025).
• Cursor: A transparent AI coding assistant where users see and approve diffs in real time.
• Bolt.new: A browser-native environment that deploys entire apps from natural prompts (Refine.dev, 2025).

We’re not just seeing hobbyist momentum here. Inside enterprises, this shift is already reshaping internal velocity. Business users are shipping prototypes, integrations, and internal tools in hours....not weeks....without ever touching a line of code.

It’s not a threat to engineering. It’s leverage. Execution is decentralising and accelerating.

Agentic Architecture & Orchestrated Workflows

Once you’ve built something, the next question isn’t “does it work?” it’s “how does it connect?”

This is where execution breaks or compounds.

We’ve entered the era of agentic architecture: AI-native systems that take a goal, break it into parts, and assign each part to a specialised agent, coordinated, parallel, and autonomous.

• One agent pulls data from your CRM or product database.
• Another drafts content or analysis based on that data.
• One validates against compliance or brand guidelines.
• Another pushes the result into Slack, HubSpot, or Notion instantly.

This isn’t just automation. It’s orchestration, with awareness of sequence, logic, dependencies, and outcomes.

Orchestration layers like LangChain (for chaining tools and memory), MindStudio (for no-code agent design), and Zapier, Make.com, and Flowise (for low-code automation) are making this composable by design (Joyce Birkins, 2025).

What used to take a product team, an engineer, and a delivery roadmap is now handled by a network of agents with your intent as the trigger.

This is where execution scales without bureaucracy.

Model Context Protocol: The Standard for Execution

Orchestration unlocked intent-based workflows. But without a shared foundation, every AI system still spoke its own language.

That’s where the Model Context Protocol (MCP) comes in.

Developed by Anthropic, MCP is the emerging connective tissue for AI ecosystems. Think of it as the USB-C of AI tooling: one universal interface that lets models discover, use, and coordinate tools, data, and actions natively.

Instead of bespoke APIs and rigid integrations, MCP introduces a common server layer exposing:

• Tools: like send_email or query_database
• Resources: such as files, customer records, or policies
• Prompts: structured templates for consistent execution

This allows agents to maintain memory, select tools mid-task, and share execution context across sessions, platforms, and models.

Early adopters include Claude, Block, Zed, and Codeium, with more layering in as interoperability becomes essential, not optional (Philschmid, 2025).

MCP isn’t just about plug-and-play AI. It’s about building a networked execution layer where agents, tools, and platforms work as one.

The AI Operating Systems: Where Tools Become Teams

When orchestration, vibe coding, and context protocols converge, we don’t just get better AI tools, we get a new class of operating system.

These are full-stack AI execution environments, built to take goals, not just instructions. Designed to deliver outcomes, not just assistance. And they’re already working in the wild.

Manus: The Agent Workforce

Manus is a cloud-native AI OS built around agent-first architecture. You assign a task like “analyse 500 CVs and generate a shortlist” and Manus handles the orchestration.

No prompt loops. No micro-managing. Each part of the job is handled by a specialised agent: coding, summarising, browsing, formatting. They work in parallel. They coordinate autonomously. And they deliver.

This isn’t a prototype. It’s a workforce.

Manus was hired on Upwork and Fiverr, completed jobs, generated deliverables, and got paid (WorkOS, 2025). It’s not pitching capabilities. It’s operating in marketplaces.

Genspark.ai: The Generalist Super Agent

Genspark takes a different path to the same destination: a mixture-of-agents model built on LLMs, integrated tools, and real-time orchestration.

Its Super Agent executes full workflows using over 80+ specialised tools—from travel planning to restaurant booking, video generation, voice calling, and even animated content creation.

It doesn’t stop there:

• Plans and books 5-day travel itineraries using map and research tools
• Generates videos from recipes or trending news with voiceovers and sound effects
• Supports marketers, teachers, analysts, and recruiters with fully packaged tasks

Why does it work? Because Genspark combines:

• Large Language Models
• Toolsets (for real-world action)
• Datasets (for nuance and context)

Together, these make it fast, reliable, and steerable—ready to execute across everyday knowledge work.

The Bigger Signal

What both Manus and Genspark.ai show isn’t competition, it’s convergence.

They represent the next phase of AI delivery: not assistants helping operators, but systems that become the operator.

For Zero-to-Solve thinkers, these platforms are not about replacing people, but about rethinking who (or what) delivers value in your execution model.

Zero-to-Solve: The Operating Model of Now

Let’s connect the dots.

The platforms. The agents. The protocols. The citizen developers. All of it is converging toward one undeniable shift:

We’re no longer building tools. We’re building ecosystems.

And ecosystems don’t just scale, they compound. Each new agent, workflow, or integration increases your organisational surface area for execution. AI becomes less of a layer, and more of a substrate, something your operations are built on.

If your delivery model still depends on:

• Manual prioritisation
• Velocity bottlenecks
• Backlog gatekeeping

…then you’re not building for capability, you’re building for delay.

It’s time to redesign the way work gets done.

Final Word: Own the Execution Layer

So here’s the question every leader should be asking:

Because the execution layer has moved. And those who see where it went will own what comes next.

Let's Cut Through the Noise

At LuminateCX, we help leaders:

• Separate signal from hype
• Identify execution leverage points
• Build AI-native workflows and Zero-to-Solve roadmaps that actually ship

Let’s design your execution model for what’s real, not just what’s possible.

🧾 References

• Garg, J. (2025). Vibe Coding: Concept, Workflow, AI Prompts, Tools. Medium.
• Codeium. (2025). Windsurf Editor. codeium.com.
• Refine.dev. (2025). Bolt.new – AI App Builder. refine.dev.
• The Prompt Warrior. (2025). Bolt vs. Cursor. promptwarrior.com.
• Joyce Birkins. (2025). 16 AI Workflow Platforms. Medium.
• Philschmid, P. (2025). MCP Overview. philschmid.de.
• Anthropic. (2024). Introducing the Model Context Protocol. anthropic.com.
• WorkOS. (2025). Introducing Manus. workos.com.
• AI Base News. (2025). Genspark Super Agent. aibase.com.

Introduction

It’s hard to avoid the buzz. If you’re a head of marketing or digital leader in 2025, you’ve probably seen more AI webinars, thought pieces, and platform updates in the last six months than in the previous five years combined.

But as the noise ramps up, many leaders are left with more questions than answers: What should we actually be doing? Are we behind? Is AI really going to replace jobs, or is it just another overhyped trend?

In this conversation, I sat down with my Business Partner Dan Shaw as he explored his questions for me, not from the perspective of a tech evangelist, but from a grounded, business-first lens. Here’s what we uncovered.

The AI Conversation Isn’t Just About AI

Despite the flood of AI content, the biggest shifts aren’t just technological. As I shared early on:

Over the last five years, we’ve seen an accelerated evolution in how businesses operate from hybrid work models to restructured teams and pressure to do more with less. AI has become part of that story, but not the whole story.

Fear vs Focus: The Right Mindset for AI Adoption

Let’s get one thing clear: the fear is real, but it’s misplaced.

It’s not wrong to feel uncertain. Budgets are tightening. Teams are shrinking. And the noise around AI is relentless. But fear shouldn’t paralyse action.

2024 was the buffer year. In 2025, the window to wait and see is closing.

AI Is Not a Silver Bullet

A critical theme we unpacked was this: AI will not save you if you don’t know what problem you’re solving.

Dan shared a scenario typical for many leaders today: where a head of marketing is grappling with reduced budget, fewer people, and constant pressure to deliver more. Vendors are knocking. AI features are everywhere. But what do you actually do?

Just like rolling out a CRM or ERP, adopting AI requires understanding:

• Your Data & internal processes
• Your team’s IP, and how to nurture & apply that knowledge
• Your brand voice & culture fit both internal & external
• Your customer nuances, and CX expectations. 

AI is powerful, but it’s only as good as the data, structure, and context you give it.

Strategy First, Always

Too often, AI is treated as the strategy. But as I emphasised during the chat:

In your business today, are you trying to automate parts of your workforce, speed up customer response times, generate better marketing content, or improve data visibility and decision-making? Until you define that, any AI initiative is just a shot in the dark.

Tap Into Existing Momentum

One of the most overlooked opportunities in early AI adoption? Your own people.

There are already individuals inside your organisation quietly using AI tools to speed up tasks, improve work quality, or test ideas. Rather than fearing shadow IT, bring these people in.

This doesn’t require enterprise-wide investment, just curiosity, frameworks, and an appetite for experimentation.

Apply What You Already Know: CRM Lessons for AI

We often think of AI as something entirely new. But the blueprint for adoption already exists.

Think back to your CRM transformation. You didn’t just install software, you defined processes, cleaned up data, trained users, and built trust across departments. AI should be no different.

It’s not about reinventing the wheel. It’s about reapplying proven practices to a new wave of capability.

What Comes Next: AI as the New Operating System

Perhaps the most forward-looking part of our conversation was this: AI is moving from ‘nice to have’ to ‘core business infrastructure.’

This doesn’t mean every business needs to rush out and build custom models or replace their SaaS stack. But it does mean we’re seeing a shift from AI as a bolt-on tool to AI as an integrated enabler across marketing, operations, customer service, and more.

Key Takeaways

Final Thoughts

This isn’t about panic. It’s about progress. AI isn’t here to take your job, but it might just accelerate your future, if you’re willing to engage with it intentionally. It’s a moment of opportunity for those who choose to lead, not just react.