Pelican Concepts

THE TAKE

Three numbers should reset your AI planning this week. Claude Sonnet 5 became the default model for every free and paid user this week, at introductory pricing below its predecessor — the AI tool your team already has open in a browser tab just got noticeably better and cheaper, whether you budgeted for it or not. Publicis Sapient’s new enterprise survey found 79% of organisations still can’t get AI adoption right, with only 18% fully integrated, even though AI already touches 53% of enterprise work. And Anthropic’s most capable model yet, Fable 5, is back in general release after a brief government-mandated suspension — proof that the frontier tools available to a $5M–$30M business today are sharper than they were a month ago, policy permitting.

Read those together and the picture is stark: the frontier keeps getting cheaper and more capable, while most companies — including plenty in Kuala Lumpur and Singapore — are stuck running pilots instead of production. That gap is the opportunity. Malaysia’s RM5.9 billion national AI push and Singapore’s new National AI Impact Programme are both explicit bets that SME and mid-market firms, not just tech giants, will capture the productivity dividend.

The tools are cheaper and more capable than they were a quarter ago. The money is flowing to whoever moves first. The window to be the AI-fluent operator in your sector — rather than the one still evaluating vendors in 2027 — is closing faster than most CEOs realise.

1. IN THE NEWS – MODEL & PRODUCT RELEASES 

Anthropic Makes Claude Sonnet 5 the Default — and Cuts the Price

Anthropic shipped Claude Sonnet 5 on June 30 and made it the default model for every Free and Pro user starting July 1. It performs close to flagship Opus 4.8 on many tasks and ships with a 1 million token context window as standard. Through August 31, introductory pricing puts it below what Sonnet 4.6 cost.

For a $10M–$100M operator, this is the real story: the “good enough” model tier just got dramatically better and cheaper at the same time. Workflows you shelved six months ago because the model wasn’t reliable enough, or the per-token cost didn’t pencil out, are worth re-testing this month.

Practical move: audit any AI pilot you paused in the last two quarters for cost or quality reasons. Re-run it against Sonnet 5 before you write it off for good.

Source: Build Fast With AI — AI News Today, July 1, 2026

Anthropic’s Most Capable Model Yet Is Back — Claude Fable 5 and Mythos 5

Anthropic released Claude Fable 5 — the first public version of its next model class, code-named Mythos — on June 9, alongside Claude Mythos 5 for vetted enterprise partners. Days later, a US export-control directive forced a sudden suspension of both models for foreign nationals, including Anthropic’s own foreign-national staff. The Department of Commerce reversed that decision on June 30, and Fable 5 returned to every user worldwide on July 1.

For a business owner, the technology is the least interesting part of this story. The interesting part is that a change in Washington policy took your AI vendor’s flagship product offline for you personally, with zero notice, for close to three weeks — even though you had nothing to do with the underlying dispute. If any part of your operations leans on a single AI vendor for something mission-critical, this is the argument for a documented fallback: a second model or vendor you can switch to within a day, not a quarter.

On capability, Fable 5 handles software engineering, knowledge work, and vision tasks, with hard safety limits that route high-risk requests — cybersecurity, biology, chemistry — to the more conservative Claude Opus 4.8 instead.

Source: Fortune — Anthropic’s Fable and Mythos Models Are Back, July 2, 2026

OpenAI Previews GPT-5.6 — But You Can’t Use It Yet

OpenAI unveiled its next model family on June 26: Sol (flagship), Terra (balanced), and Luna (fast, low-cost). Access is currently gated to roughly 20 trusted organisations, with general availability planned for “coming weeks.”

For mid-market leaders, this is a heads-up, not an action item. Don’t build a Q3 roadmap around a model you can’t yet buy. It does confirm that tiered pricing — flagship, balanced, cheap — is becoming the industry standard, which means your AI vendor costs should keep dropping across all three tiers as this rolls out.

Watch for general availability over the next few weeks — that’s when it becomes relevant to your procurement conversations.

Source: LLM Stats — LLM News Today, July 2026

Google Pushes Gemini 3.5 Pro to General Availability

After slipping from a June target, Gemini 3.5 Pro is cleared for general availability this month, positioned against Gemini 3.1 Pro, currently one of the top-ranked reasoning models on independent benchmarks. Google also shipped Gemini 3 Pro Image, aimed at design and marketing production work.

If your business runs on Google Workspace, this matters more than the headline suggests. Google is embedding Gemini directly into every Workspace surface, so the upgrade may land inside your existing subscription with no new procurement decision required.

Check with your IT lead or Workspace admin on rollout timing before assuming you need a separate AI budget line for this one.

Source: AI Release Tracker — Latest AI Model Releases, July 2026

2. ENTERPRISE ADOPTION & CASE STUDIES

97% of Executives Deployed AI Agents. Only One in Five Can Govern Them.

Publicis Sapient’s 2026 Global Enterprise AI Report found AI now supports 53% of enterprise work, yet only 18% of organisations call themselves fully integrated. Separately, 97% of executives say their company deployed AI agents in the past year — but only one in five have a mature governance model for them.

“54% of C-suite executives admit AI adoption is straining their organisation internally.”

The lesson for a mid-cap operator: deployment speed without governance is now the median failure mode, not the exception. If you’re rolling agents into finance, customer service, or operations, put a named owner and a review cadence in place before you scale past pilot — that’s the difference between the 18% who are fully integrated and the 79% still fighting their own rollout.

Source: Publicis Sapient — 2026 Global Enterprise AI Report

Telecom and Retail Are Pulling Ahead on Agentic AI

Telecommunications firms lead agentic AI adoption at 48%, with retail and consumer packaged goods close behind at 47%, per new industry survey data. Worker access to AI tools inside organisations rose from under 40% to roughly 60% in a single year.

If your sector isn’t telecom or retail, treat this as an early-warning signal, not a reason to wait. The adoption curve that started in those industries eighteen months ago is now showing up in professional services and manufacturing. Movers in year two of a sector’s AI curve capture most of the advantage; movers in year four are playing catch-up against entrenched competitors.

Source: NVIDIA — State of AI Report, 2026

Users Are Ditching “Tokenmaxxing” for Efficiency

A growing share of OpenAI’s and Anthropic’s own customers are shifting from maximising token usage to optimising for cost efficiency, as enterprise AI adoption matures past the “throw compute at everything” phase, according to reporting this week.

For a mid-market operator paying by usage, this is good news: it signals vendors are being pushed toward more cost-disciplined pricing and tooling by their own biggest customers. Before renewing any AI contract, ask what efficiency-mode or reduced-cost options now exist that didn’t six months ago — you’re negotiating from a stronger position than you were in Q1.

Source: CNBC — OpenAI and Anthropic Face New AI Reality, June 26, 2026

3. FUNDING & M&A

AI Funding Hit $510 Billion in H1 2026 — 43% Went to Two Companies

Global venture funding reached a record $510 billion in the first half of 2026. OpenAI and Anthropic alone accounted for $217 billion of that — 43% of all startup capital raised worldwide. Nearly 88% of all AI funding went to US-headquartered companies.

Practical implication for a mid-market buyer: this concentration lets the frontier labs subsidise pricing and outspend smaller vendors on distribution — good news for your input costs, bad news if you’ve bet your stack on a smaller, undifferentiated AI vendor. Ask any vendor with under $50M raised how they plan to compete on model access and pricing over the next 18 months.

Source: Crunchbase News — Record-Breaking AI Funding, 2026

Together AI Raises $800M; Anthropic Buys Its Way Into Biotech

Together AI closed an $800 million Series C led by Aramco Ventures, one of the largest infrastructure raises of the year. Separately, Anthropic acquired computational biology startup Coefficient Bio for roughly $400 million in an all-stock deal and hired AlphaFold lead John Jumper away from Google DeepMind.

Neither deal touches your business directly, but both signal where frontier labs are placing bets beyond chat: compute infrastructure and vertical science applications. If you’re in healthcare, pharma, or a life-sciences-adjacent industry, expect the first serious AI-native competitors in your space to emerge from exactly this kind of lab-to-vertical move within 12–18 months.

Source: Tech Startups — Venture Capital & Startup Funding Roundup, July 1, 2026

4. SOUTHEAST ASIA SPOTLIGHT — MALAYSIA & SINGAPORE

Malaysia Commits RM5.9 Billion to AI — and Locks Data Centres to AI-Only Use

Prime Minister Anwar Ibrahim’s 2026 budget allocates RM2 billion (~US$490 million) to build a sovereign AI cloud, letting Malaysian organisations train and deploy models on domestic infrastructure rather than relying solely on foreign-owned cloud providers. It sits inside a wider RM5.9 billion (~US$1.44 billion) commitment to AI research, development, and commercialisation. Malaysia has also stopped approving new data centre projects that can’t demonstrate an AI or high-tech use case — a direct policy signal about where the government wants capital to flow.

RM5.9 billion (~US$1.44 billion) — Malaysia’s total 2026 commitment to AI research, development, and commercialisation.

For a Malaysian CEO, the practical read is this: sovereign infrastructure and RDCI grants are coming online this year. If data residency, compliance, or cost has been your reason to delay a serious AI infrastructure investment, policy is actively dismantling that objection. Get your team in front of MOSTI or MDEC now, before the RDCI funding queue gets long.

Source: MyDIGITAL — Budget 2026: Accelerating Malaysia’s Digital Transformation

Singapore’s Budget 2026 Puts Real Money Behind SME AI Adoption

Singapore’s Budget 2026 introduced a 400% tax deduction for AI investment (capped at S$50,000 per assessment year), an expanded Productivity Solutions Grant covering more AI-enabled tools with up to 50% co-funding, and a new National AI Impact Programme targeting 10,000 enterprises and 100,000 workers over three years. The flagship Champions of AI Programme, for companies ready to move beyond pilots, launches later this year. Forecasters have also raised Singapore’s 2026 GDP outlook to 4.3%, citing sustained AI-related investment alongside a construction boom.

For an SME or mid-cap CEO in Singapore, the arithmetic changed this quarter: co-funding plus a tax deduction plus a skills subsidy means your effective cost of a serious AI adoption programme is meaningfully lower than it was six months ago. If you haven’t mapped your planned AI spend against these three schemes, that’s a half-day exercise with a direct payback.

Source: Singapore Budget 2026 — Harness AI as a Strategic Advantage

5. TOOLS & WORKFLOWS WORTH TRYING

Claude Sonnet 5 — Now the Free-Tier Default, Worth a Second Look

What it does: Anthropic’s most agentic Sonnet model yet, with a 1 million token context window and pricing below the previous generation through August 31.

Use case for a CEO: point it at a genuinely messy multi-step task you’ve been doing manually — reconciling a vendor contract against your actual invoices, drafting a first-pass board deck from raw operating data, or triaging a backlog of customer emails against your policy documents. The larger context window means it can hold an entire contract or a quarter’s worth of data in a single pass instead of you chunking it manually.

Source: Build Fast With AI — AI News Today, July 1, 2026

Zapier’s AI Copilot — Build Automations by Describing Them

What it does: Zapier now bundles Tables, Interfaces, and Zapier MCP into standard plans, and its AI Copilot builds multi-step automations from a plain-language description. Its Agents feature creates assistants that take multi-step actions across your existing tool stack.

Use case for a CEO: describe a recurring admin task in plain English — “when a new lead enters our CRM, check it against our disqualification criteria, and if it passes, schedule a call and notify the sales lead on Slack” — and let Copilot build the workflow instead of briefing a developer. It’s the fastest path to a working automation without adding headcount or a dev queue.

Source: Zapier — Best AI Productivity Tools, 2026

THIS WEEK’S TAKEAWAY

Ninety-seven percent of executives say they deployed AI agents this past year. Only one in five say they can actually govern what those agents do. If your competitors are moving faster than their ability to control what they’ve built, is the smart move to move faster too — or to make sure your own foundation can hold the weight before you add another agent to it?

EVENTS

AIBP C&E Kuala Lumpur — July 8–9, 2026. 600+ enterprise leaders on AI transformation, cybersecurity, ESG, and digital workforce. aibp.sg/conference-exhibition-malaysia

Conversational Tech Summit Asia — Kuala Lumpur, Singapore & Jakarta, 2026. CEOs and practitioners on generative and agentic AI for customer experience. conversationaltechsummitasia.com

📬 Subscribe to the Weekly AI Digest Each Monday morning, Peter Lam distils the week’s most actionable AI developments for SME and mid-cap CEOs — no hype, no filler, just what matters for your business. Subscribe at peterlamcoach.com · Forward to a CEO who should be reading this.

Pelican Concepts  |  peter@peterlamcoach.com  |  peterlamcoach.com  |  Week of 23-29 June 2026