The tech media has moved on from AI model performance, at least according to Brian Madden's recent Citrix blog post. He argues that the real AI race isn’t happening in the labs, but in the workplace, where productivity, trust, and adoption are the new battlegrounds.
That view is largely true, but also incomplete.
The Heart of Brian's Argument
Brian's key point is this: AI is no longer just about the models. It’s about how organizations adopt and integrate them. From document drafting to real-time support, the way AI reshapes day-to-day workflows is where the real competitive edge lies.
"The future of work is blended, not hybrid," he writes, highlighting that human-AI collaboration will become the new normal, and that success lies in operationalizing AI, not marveling at its capabilities.
Why This Is Right, Mostly
- Workflows are already changing. Teams are using AI to co-draft proposals, write code, and analyze documents.
- The cultural gap is real. Many organizations are held back not by a lack of tech, but by a lack of trust, training, or structured adoption.
- Blended work is the future. AI isn’t replacing workers. It’s partnering with them. The winners are those who adapt early and well.
But here’s where we need to push back a little.
The Model Still Matters, and Here’s Why
To say the model wars no longer matter is to overlook a key truth: the model defines the ceiling of what your AI can do.
Consider:
- Privacy and sovereignty: A business using GPT-4 may face different regulatory or cost challenges than one using open-source Mistral or LLaMA.
- Hallucination and trust: A poor model undermines adoption by generating untrustworthy results.
- Use-case fit: Legal, financial, or multilingual tasks demand high-quality, often fine-tuned models.
You can’t build great AI-enabled workflows on weak foundations. Model performance still matters. It just isn't the only thing that matters.
The Bigger Picture: Stack, Not Spotlight
The future of AI at work isn’t a choice between "model or workflow." It’s about a stack:
- Strong foundational models
- Clear orchestration (RAG, agents, workflows)
- Seamless human-AI collaboration
- Trust and governance frameworks
- Continuous feedback loops for improvement
Each layer shapes the others. Ignore one, and the whole thing wobbles.
Let’s Be Honest: Some Jobs Will Disappear
While it’s encouraging to talk about "blended work" and human-AI collaboration, we also need to be brutally honest:
Some categories of workers will be replaced by AI. Not augmented. Not supported. Replaced.
Customer support roles that follow rigid scripts, data entry jobs, junior analysts who compile dashboards, transcriptionists, low-complexity paralegal work. These are already being disrupted. Pretending otherwise is not optimism. It’s denial.
Organizations that ignore this reality are setting themselves up for workforce shocks. And employees who don’t actively reposition themselves to work with AI may find themselves competing against it.
We need a dual mindset: yes, invest in AI-human collaboration. But also prepare for the socioeconomic shifts that full or partial automation will bring. Ethics and empathy must be part of the deployment strategy, but they cannot shield us from structural change.
My thoughts
Brian Madden is right to urge us to move beyond the hype. But let’s not throw out the core of what makes AI powerful.
Yes, adoption, trust, and workflow design are where real transformation happens.
But let’s not pretend the model no longer matters. It still defines what’s possible.
As we move forward into the age of AI-augmented work, the question isn’t "model or workflow?" It’s how we align both to build something useful, ethical, and scalable.
Inspired by, but respectfully divergent from, Brian Madden’s blog post on Citrix.
