This issue is about AI agents going to work for real. Globally, agents are starting to do the tasks people used to do inside software — enough that Gartner says a fifth of the SaaS market is up for grabs — while AI spending jumps 63% in a year and finance teams start asking harder questions about value. In the Gulf the move is sharper still: Ajman just renewed a trade licence with an AI agent, no human in the loop, and Saudi's HUMAIN is building sovereign, Arabic-language AI with Cohere. The pattern underneath it all is the one TASC keeps seeing: the technology is the easy part now. Getting value from agents — safely, with a person still accountable — is a people-and-governance job. Here's the news, why it matters, and a playbook to get there.
What's moving in AI worldwide
AI agents are quietly breaking how software makes money
AI agents are starting to do the work people used to do inside software — and that quietly breaks how a lot of enterprise software gets paid for.
$234bn — of enterprise software spend exposed to agentic AI by 2030 (~20% of SaaS) (Gartner)
For years, buying enterprise software meant buying seats and screens. Agentic AI changes that. When an agent handles a whole task across several systems, the interface people used to pay for becomes optional — and Gartner estimates up to $234 billion of application spending, about a fifth of the SaaS market by 2030, is exposed to this shift.
This isn't the end of software; it's a change in what you pay for. Gartner's word for it is the "Saaspocalypse" — the old seat-license model breaking up as buyers ask for outcomes instead of more features and more dashboards. Less an apocalypse, in its words, than a metamorphosis.
The opening this creates is for whoever can turn AI into a measured result and run agents safely across the systems you already have. That's less about owning another tool and more about the people, process and governance wrapped around it — which is exactly what the rest of this edition is about.
Why it matters: Value is moving from software you log into to agents that just get the result. The winners will be the providers and partners who can deliver measured outcomes and run agents safely across your existing systems — not the ones defending dashboards.
Source: Gartner — $234 Billion in Enterprise Application Software Spend Is at Risk from Agentic AI
AI spending is up 63% — but the money now chases value, not hype
Companies are spending far more on AI this year, and they've started asking a much harder question: what am I actually getting for it?
63% — jump in global AI model & platform spending in 2026, to $64bn (Gartner)
The money is pouring in. Gartner expects spending on AI models and platforms to reach $64 billion in 2026, up more than 63% in a single year, with generative-AI model spend more than doubling.
But the mood has changed. Finance is watching. Gartner says budgets are now under scrutiny and buyers want usage efficiency, cost control and outcomes they can measure — and the fastest-growing slice is specialised, domain-specific models (up 210%), not general chatbots.
The caution behind the boom: Gartner still expects more than 40% of agentic AI projects to be cancelled by the end of 2027, mostly over cost, unclear value and weak controls. The takeaway for leaders is simple — spending more is easy; getting value is the actual job. That's a discipline-and-governance problem, which is where this edition's playbook comes in.
Why it matters: More spend doesn't mean more value. The leaders this year treat AI as a governed, measured programme — with clear outcomes and cost control — not a shopping spree. The fastest-growing spend is on specialised, domain-specific models, not general chatbots.
Source: Gartner — Worldwide AI Platforms and Models Market to Grow 63% in 2026
AI in the region: UAE & KSA
The UAE just ran a real government service on an AI agent
The UAE isn't piloting agentic AI in government — it's starting to run live public services on it, with no human pushing the buttons.
50% — of UAE government services targeted to run on AI agents within two years (UAE Cabinet)
Ajman has done something most organisations are still only talking about: it renewed a trade licence with an AI agent, start to finish, with no one pushing the buttons. The system notices the licence is expiring, walks the customer through the steps in the AjmanOne app, pulls in the municipality when a lease needs checking, and completes the renewal — a "proactive, headless" government service.
It's the first live example of a much bigger push. In April the UAE Cabinet set a target for AI agents to deliver half of all government services within two years, and the country plans to train 80,000 government employees to work alongside them.
For any employer in the UAE, that resets expectations. If the government can run a regulated service on an agent — with, as Ajman's leadership put it, governance, accountability and data protection built in — customers and regulators will expect the same standard from you. The message isn't "experiment"; it's "deploy, and govern it properly."
Why it matters: When the government runs live, regulated services on agents and trains tens of thousands of people to supervise them, AI fluency stops being optional for anyone operating in the UAE — and the bar for doing it with proper governance rises just as fast.
What to do: Treat agentic AI as a near-term reality, not a 2027 project. Pick one high-volume, rules-based process, put an agent on it end to end with a named person accountable for the outcome, and start building the AI-supervision skills your team will need.
Source: The National — Ajman uses agentic AI to renew trade licence in UAE government first
Saudi's HUMAIN teams up with Cohere to build sovereign, Arabic AI
While others buy AI, Saudi Arabia is building its own — and just brought in a global model-maker to help.
50MW — of compute HUMAIN is dedicating to Cohere's models for sovereign AI (HUMAIN / Cohere)
Saudi Arabia's approach to AI is to own the stack, not rent it. This fortnight its national AI champion HUMAIN, backed by the Public Investment Fund, signed a strategic partnership with Canada's Cohere: at least 50 megawatts of dedicated compute for Cohere's models, and a joint effort to build enterprise and sovereign AI — including Arabic-language models.
"Sovereign" is the key word. It means models built, hosted and controlled in-Kingdom, tuned for Arabic and local rules — HUMAIN already has its own Arabic model, Allam, and this deal adds enterprise muscle. For Cohere it's a first major move outside North America.
For businesses in Saudi, the direction is clear. Regulated work will increasingly run on AI that keeps data in-Kingdom and works properly in Arabic. Planning for that now — the data, the language quality, and the local talent to govern it — will matter more than which chatbot you picked.
Why it matters: "Sovereign AI" — models built, hosted and controlled in-Kingdom, tuned for Arabic and local rules — is becoming the expectation for regulated Saudi work in government, banking and healthcare. Data residency and Arabic-first AI are heading toward being the default, not a special request.
What to do: If you operate in Saudi, assume sovereign, Arabic-capable AI is where regulated workloads are going. Prioritise use cases that keep data in-Kingdom, build Arabic-language quality into anything you deploy, and line up the people who can run and govern it locally.
Source: Arab News — Saudi Arabia's Humain, Canada's Cohere join forces on AI infrastructure
AI agents & your workforce
How to put an AI agent into your hiring without losing the human touch
Agents are moving into everyday work, and hiring is one of the first places most teams feel it. Here's how to add one without losing the human judgement a good hire depends on.
Five steps to agent-assisted hiring that a recruiter still owns
- Start with the CV pile, not the decision. Point the agent at first-pass screening against clear, job-related criteria. It's high-volume, lower-risk and easy to measure. A tool like Hyrra.ai does exactly this and cuts time-to-shortlist.
- Write the criteria down first. An agent is only as fair as the rules you give it. Agree the must-haves and nice-to-haves up front, in plain language, before it reads a single CV.
- Keep a recruiter on every shortlist and rejection. The agent ranks and recommends; a person decides who moves forward and who gets a no. That human sign-off is the point, not a nicety.
- Check for bias, then check again. Review who the agent is filtering in and out. If a group keeps dropping off, fix the criteria — don't wait for a complaint.
- Measure time saved and quality. Track time-to-shortlist next to quality-of-hire and offer-accept rates. Speed only counts if the hires are as good or better.
An agent should give your recruiters their time back, not take their judgement away. Set the rules, keep a human on every call, and you get faster hiring you can still stand behind.
How to put your first AI agent into a real workflow — safely
Gartner expects more than 40% of agentic AI projects to be cancelled by 2027 — usually over cost, unclear value and weak controls. Here's how to launch your first agent so it lands in the winning 60%.
Five steps from demo to a governed agent in production
- Pick a boring, high-volume process. Not your flashiest use case. Choose something repetitive and rules-based where success is easy to measure, and prove it there first.
- Redesign the work around the agent. Don't bolt an agent onto an unchanged process. Map where it acts, where it hands off, and where a person steps in — that redesign is where the value is.
- Name a human owner for every decision that matters. Someone must be accountable for what the agent does. Give routine steps light oversight; anything touching money, customers or compliance needs a person who can see and override it.
- Log everything and set hard limits. Keep an audit trail of what the agent did and why, and cap what it's allowed to do. If you can't explain a decision afterwards, you're not ready to scale it.
- Put a number on value before you expand. Agree the cost, time or quality gain up front and check it's real before rolling wider. No clear number is exactly how projects end up cancelled.
Agents fail on governance and value, not on technology. Start small, keep a human accountable, log the lot, and only scale what you can prove — that's the difference between a pilot and a payoff.
The Leader's Playbook for Putting AI Agents to Work
Playbook. Every leader is being told to "deploy agents." Few are told how to do it without landing in the 40% of projects that get cancelled. This six-step playbook walks C-suite, Directors and VPs in UAE and KSA enterprises from a first agent to governed, scaled, value-generating agentic AI: where to start, how to redesign the work, how to build governance and human accountability in from day one, how to meet regional expectations like sovereign and Arabic-first AI, and how to close the AI talent and supervision gap without out-bidding the whole market.
Spotlight: AI-Assisted Hiring with Hyrra.ai
Hyrra.ai. Hyrra.ai puts AI agents to work on the CV pile — automating first-pass screening and cutting time-to-shortlist, with a recruiter in control of every shortlist and rejection. It's agent-assisted hiring done the way this edition recommends: the agent clears the volume, your people own the decisions.
