Performbyai

Consulting for Strategy Execution, AI Transformation, and Market Intelligence. Built to deliver at a speed and value few consulting firms can match.

[Performbyai @performbyai]

https://t.co/DqOp82JxPC

[Performbyai @performbyai]

In 1982, behavioral economists proved a rule of human nature that dismantles how most businesses operate.

It is called the Ultimatum Game.

Person A is handed $100 and must propose a split with Person B.

If Person A offers $10 and keeps $90, Person B has a choice: accept the $10, or reject it.

If rejected, the money vanishes. Neither gets a dime.

Pure economic logic dictates taking the $10. Any profit is better than zero. But human nature dictates otherwise. People overwhelmingly reject the unfair split. They will willingly take zero just to ensure the party hoarding the value also loses.

The lesson is absolute: attempting to extract all the value for yourself breaks the entire system.

Yet, this is exactly how traditional consulting models are built. Firms operate in walled gardens. They hoard clients and capabilities and work in isolation. The result is ultimately slower growth and eventual decline.

At Perform by AI, our architecture is built on the exact opposite behavioural truth.

We chose to operate on trust, shared values and fairness with the partners we collaborate with, and the clients we deliver for.

That's what we mean when we say Maximum Value.

Decide to be unfair. Lose everything.
Collaborate and share. Gain everything.

#performbyai

[Performbyai @performbyai]

We are proud to introduce our Senior Advisory Board at Perform by AI.

The role of the board is to reinforce our commitment to building trust and delivering sustainable impact in everything we do. Each bringing with them deep expertise, and a shared belief in strong ethics and values.

𝗗𝗿 𝗛𝗶𝘀𝗵𝗮𝗺 𝗔𝗯𝗱𝗲𝗹 𝗥𝗮𝗵𝗺𝗮𝗻
Leadership & Governance
Led healthcare organisations at board and chief-executive level across the UK and Asia, including an NHS hospital and a health system he built and governed in Shanghai.

𝗣𝗮𝘂𝗹 𝗚𝗼𝘀𝗻𝗲𝗹𝗹
AI Automation & Integration
Founded and led several technology and product companies across the GCC, UK and US, building AI-native delivery models and digitisation programmes for government and enterprise.

𝗬𝘂𝘀𝗿𝗮 𝗔𝗵𝗺𝗮𝗱
Ethical AI & Workforce Upskilling
Led Data & AI transformations for FTSE 500s and UK/GCC government entities. Partnered with top UK universities on responsible AI capability development. RED Foundation GCC Chair, published author, GIRAI Country Researcher.

𝗗𝗿 𝗣𝗮𝘆𝗺𝗮𝗻 𝗦𝗮𝗱𝗲𝗴𝗵
AI Data Analytics & Machine Learning
Served as chief data officer in the US federal government and at Cushman & Wakefield, a former White House Presidential Innovation Fellow who holds machine-learning patents.

Having leaders of this standing for our clients across the GCC region is a deliberate choice about the company we intend to become.

#performbyai #advisoryboard #responsibleai

[Performbyai @performbyai]

By 2028, Gulf businesses measured by cost cuts or AI spend alone will have missed the more important conversation about how they protected margin.

The four levers, sequenced by cash position, protect margin under pressure.

The key takeaways are in the carousel below.

Read the full article: hhttps://x.com/performbyai/status/2059891519134511486?s=20

#performbyai #costpressure #gulfbusiness

[Performbyai @performbyai]

Under pressure, one move is never enough.

Cost pressure is the new floor, not a phase. The two instinctive answers both fail: cut alone and you shrink, buy technology alone and you stand still.

1 | RECLAIM TIME. Most teams run at 50 to 60% effective, so the first gain is the hours you already pay for, ahead of any new technology. Managerial discipline alone returns 10 to 20% inside 90 days, with no significant technology spend.

2 | REDIRECT SPEND. Around 70% of the technology budget funds what already exists. The leaders retire what no longer earns its place and send the savings to the work that does. Reallocation is a discipline, not a budget line.

3 | APPLY THE RIGHT TOOLS. Narrow beats broad, every time. Pick a few workflows where the data is good and the impact is measurable, redesign them end to end, and tie every result to a number. Around 20% of firms capture three quarters of the returns.

4 | REACH FURTHER. The capacity the first three moves free up is for the work you could not take on before, a new segment, a new product, a new market. The leaders here are 2.6 times more likely to reinvent the business model.
The squeeze is already sorting firms by four results: margin protected, capital redirected, work redesigned, reach extended. AI spend will not be on the list.

Which of the four can yours show today?

Let's talk.

#performbyai #costpressure #strategyconsulting

[Performbyai @performbyai]

Our Founder & CEO Mo Smadi has published a new article on cost pressure for Gulf businesses in 2026.

The pressure is structural, not transient. UAE non-oil firms reported selling charges rising at their fastest pace in 11.5 years through March 2026, and the International Labour Organization's severe-escalation scenario projects GCC hours worked falling 11.5% and employment falling 7.1%.

The default responses fail empirically. EY's 2026 CEO Outlook finds 61% of CEOs expecting higher operating costs against only 16% expecting reductions. BCG's 2026 research shows 60% of companies report minimal or no value from AI despite significant effort.

The article sets out a four-lever model that pairs productivity, strategic reallocation, selective AI, and reach expansion, sequenced by cash position. Cash-positioned businesses run all four in parallel. Cash-constrained businesses sequence from productivity, letting savings fund the rest.

Read the article on X.

#PerformbyAI #CostPressure #GulfBusiness

[Performbyai @performbyai]

https://t.co/DqOp82JxPC

[Performbyai @performbyai]

@weareageance Thank you for sharing @weareageance

[Performbyai @performbyai]

Jobs created minus jobs lost is not a clean number, it varies by industry and by function, and even when the math comes out positive, the people losing roles are not the people stepping into the new ones. You cannot retrain everyone and companies will selectively decide who is worth retraining. Moreover, productivity and efficiency are at its core, and the cost of manual human labour is its target.

Enterprises and businesses that are looking to achieve a 20-30% improvement in productivity will have to match that with an equal level in sales and growth, which is nearly impossible to achieve for most.

Whether that becomes the kind of crisis where millions are out of work has very little to do with AI itself. It comes down to what governments fund, what regulators are willing to enforce, and what companies choose to absorb rather than push downstream.

That last part is on us, not on the technology.

[Performbyai @performbyai]

Wishing everyone a blessed Eid Al Adha.

We hope this occasion brings prosperity to your businesses, joy to your homes, and a well-deserved opportunity to rest and recharge with your loved ones.

Eid Mubarak from all of us at Perform by AI.

#EidAlAdha

[Performbyai @performbyai]

Microsoft Research just stress-tested how AI fails in real workflows. Five findings break most teams' deployment assumptions.

The benchmark, DELEGATE-52, tests 19 large language models across 52 professional domains over 20-interaction workflows. Frontier models corrupt 25% of document content. The findings underneath that headline are what teams deploying AI should be reading.

Models tested span every major frontier system: GPT, Claude, Gemini, Mistral, Grok, Kimi. Five findings rewrite what most teams assume about AI delegation.

Five findings underneath the headline

▸ The drift mental model is wrong. AI fails in catastrophic bursts. Across 19 models, 80-98% of total degradation comes from rare single-round failures, not gradual accumulation. Stronger models have fewer big errors, not smaller ones.

▸ Better models corrupt. Worse models delete. The failure mode is getting harder to detect. Frontier models like Claude 4.6 Opus and Gemini 3.1 Pro corrupt content rather than delete it. 70-78% of their errors are wrong content sitting in the right place.

▸ How a model performs after 2 interactions does not predict how it performs after 20. GPT 5 and Kimi K2.5 score 91.5 and 91.1 at two interactions, essentially tied. By 20 interactions they sit at 48.3 and 64.1. Most AI evaluation measures the wrong horizon.

▸ Agentic tooling, applied naively, makes reliability worse rather than better. Four GPT models tested with a basic agentic harness performed 3-9% worse with tools than without. Agentic does not automatically equal reliable.

▸ AI is broadly ready for coding and broadly not ready for almost anything else. Of 52 professional domains, Python is the only one where most models cross 98% reliability. Even the top model, Gemini 3.1 Pro, is ready in 11 of 52.

AI delivers in narrow programmatic domains and breaks down across most knowledge work. The failures are getting harder to detect as models improve. The most common mistake teams make is generalising from coding capability to broader knowledge work.

How is your team designing for AI failures that look like success?

Source: LLMs Corrupt Your Documents When You Delegate, Microsoft Research, April 2026.
Image: front page of the cited paper, Microsoft Research.

#LLMs

[Performbyai @performbyai]

Stanford's HAI has just shown how to tell, in real time, whether an AI is right about any specific decision it is making. Most companies running AI today have no way to check this.

Stanford HAI calls the method the Ensemble Monitoring Model. The mechanism is simple: five smaller AI models, each trained to do the same job as the primary one, give their answer in the background. The user sees a green light if they all agree with the primary AI, and a warning if they disagree. For the first time, the person using AI in production knows, in the moment, when the AI's answer is on solid ground and when it is not.

Three implications for every business now running AI in production

▸ AI is now embedded in every sector where consequential decisions get made: banking, insurance, healthcare, hiring, and the courts. In every one of those sectors, the person acting on the AI's output has no real-time signal of when to trust it. The fraud analyst clearing an AI alert, the recruiter ranking a candidate, the underwriter accepting a risk score, and the doctor reading an AI scan are all making calls without that signal. Most default to either rubber-stamping the answer, which is dangerous, or overriding it on instinct, which is wasteful. Both look like adoption from a distance. Neither one is.

▸ Every company running AI has been carrying a risk it could not measure. When an AI decision goes wrong, whether that is a wrongly denied loan, a missed fraud, or a wrong diagnosis, no one in the building can show the regulator, the customer, or the board what the AI was thinking at the moment of decision. There is no audit trail, no evidence of due diligence, and no defence. Stanford has just changed what reasonable looks like in production AI.

▸ Stanford has just set the standard for trustworthy AI in production before any regulator could. The architecture is industry-agnostic. The companies that adopt this kind of signal voluntarily will set the bar for their sector. The ones that wait will inherit it from regulators, customers, and courts that will not wait politely. The window for voluntary adoption is open now and closes the moment a regulator arrives.

The question this paper has made answerable is the one every board has been quietly worried about: is the AI we have been paying for actually right for the decisions it has been making in our business?

For the first time, the answer is something other than "we hope so."

Source: Operationalizing Real-Time Monitoring of Clinical AI, Stanford HAI, 2026
Image Credit: Stanford HAI, May 2026

hashtag#performbyai hashtag#aitrust hashtag#responsibleai

[Performbyai @performbyai]

Saudi Arabia's consulting freeze, and the truth behind the headlines.

The Financial Times reported on Thursday that Saudi Arabia has paused new contracts with western consultancies and slowed payments to end of June. The story is correct. Most of the coverage around it is reading it wrong.

Semafor expanded the FT story, and the Saudi Ministry of Finance has pushed back on parts of the payment narrative.

The facts behind the headlines

▸ The spending review is not new. Saudi has been raising the value bar on services procurement for over a year. The finance ministry began demanding more discipline on services spend through 2025, well before the war. A consulting market that grew 38% in 2022 was forecast at 13% for 2025. The freeze is the next step on a curve that has been visible for months, not a sudden retreat from Vision 2030.

▸ The payment story is more nuanced than the headlines suggest. The Ministry of Finance has contested the delay narrative, citing 99.5% of invoices paid within the contractual timeframe year to date, and more than 85% paid weeks ahead of the due date. Any consulting firm running large public sector engagements in the Gulf has long built around payment cycles that require active cash management. The exposure that matters now is on existing invoices through Q2, not on the long-run discipline of the buyer.

What every headline is missing is already in the P&L. Brent crude is trading above $104 a barrel, roughly 50% above pre-war levels, with diesel and freight up faster. Across fuel, travel, utilities, freight, and expatriate housing, the operating cost base for any consulting firm running senior teams in the Gulf has moved up structurally since February. Margins were already compressing under value scrutiny. The war has added a cost shock before a single new contract resumes.

The bigger truth is that consulting has delivered real impact in the kingdom and will be critical to the next phase. Saudi's progress under Vision 2030 in AI, tourism, non-oil growth, and digital infrastructure was built in partnership with external consulting firms. What is changing now is the buyer's bar. The market is moving from deliverables to outcomes. The firms that meet that bar will keep growing. The ones still selling decks will not.

How is your consulting firm preparing for this? Is the impact on your budget already clear?

Source: Financial Times, 21 May 2026

#saudiarabia

[Performbyai @performbyai]

Can AI solve the problem of poverty and unemployment by directly employing and paying humans to help it work ? Anyone with access to the internet, the model, and the form of receiving payment stands to benefit.

The reliance on job applications, economies, conflict, supply chains, and geographic inequality no longer become a constraint for earning a living.

[Performbyai @performbyai]

Fact 1. The drift mental model is wrong. AI fails in catastrophic bursts.

Across 19 models, 80-98% of degradation comes from rare single-round failures, not gradual accumulation.

Stronger models have fewer big errors, not smaller ones.

[Performbyai @performbyai]

Resilience is built before the crisis arrives.

Bones grow under load, and businesses sharpen under competition. Without stress, neither builds strength.

But past 85% of capacity, the same load that built the system starts destroying it. The body burns out, the factory breaks down.

And underneath both sit the vitals. For a person, sleep, food, relationships. For a business, cash, governance, values. Lose one and nothing else matters.

A resilient business does not become resilient when the shock arrives. It was built that way long before, by deliberate choice.

Which of the three is yours getting wrong right now?

𝗟𝗲𝘁'𝘀 𝘁𝗮𝗹𝗸.

#performbyai #resilience #strategyconsulting

[Performbyai @performbyai]

📑 How do you maximise impact and minimise risk from AI?

More than 80% of enterprise AI spending in 2025 produced no measurable value. The technology was almost never the reason. Everything that decides AI's impact, and its risk, happens around the technology.

Perform by AI brings strategy, governance, and the work of running AI together in one place.

Impact and risk are the same problem. They must be solved together.

𝗧𝗵𝗮𝘁 𝗶𝘀 𝘁𝗵𝗲 𝘄𝗼𝗿𝗸 𝘄𝗲 𝗱𝗼.

#PerformbyAI #AIstrategy #AItransformation

[Aegance @weareageance]

This service is a game-changer, tripling profitability and enhancing client results with faster, superior outcomes. For @performbyai, it's about delivering excellence efficiently. #BusinessGrowth #ClientSuccess

[Performbyai @performbyai]

From the Founder | Part 2: The Problems We Solve

In our latest episode, Mo Smadi, Founder & CEO of Perform by AI, lays out the two-sided problem that led him to build the firm.

The consulting market is broken on both sides.

Organisations that need real consulting and AI work face a long and expensive decision. Months get lost just choosing a partner. Fees are paid for senior names but the work is done by junior teams. The scope ends up shaped by what the firm can deliver rather than what the business actually needs. And the more the business invests, the further removed the senior expertise becomes from the actual delivery.

Consulting firms face the opposite problem. They lose millions in revenue every year on proposals they cannot put forward and tenders they cannot meet because their teams are committed elsewhere. Clients keep asking for more at lower fees, margins shrink, and teams get pulled across too many engagements at once. Growth gets capped by capacity, not by demand.

We founded Perform by AI in Dubai in 2025 to serve both sides of that market. For organisations, we deliver the full consulting and AI spectrum: strategy, performance improvement, AI transformation, market intelligence, and the expertise to execute. We do this faster and leaner, with the seniority committed to the work rather than attached to the proposal. For consulting firms, we extend the same capability into their proposals, their tenders, and their delivery, so the work they would otherwise lose gets won and delivered.

#PerformbyAI #FounderStory