Practical AI
Same work.
Less time.
Human verified.
We show your people which tasks to hand to AI, which tool suits each, and how to verify the result safely.
Same task, two ways.
Pick a task. The AI time includes a person verifying the result.
The taskTurn a 40-page policy review into a one-page briefing for the board.
Typical time
The usual way
Policy Review Update
The policy review was carried out over the summer and looked at a number of areas across the organisation. Not all findings are included here for reasons of space.
In terms of information security the review found that MFA is not enforced for all staff and recommends that this should be done, possibly by March (TBC). Supplier access is reviewed annually and could be more frequent.
Check page number?? p.12 or p.14
Remote working: see section 6 of the review for full details. Costs to be confirmed with finance.
With AI + human verification
Policy review: three decisions for the board
Enforce multi-factor sign-in for all staff by March p.4
Optional for 30% of accounts, and rated the highest risk.
Review supplier access every quarter p.12
Two former suppliers still had working accounts.
Rewrite remote-working rules for personal devices p.31
The current rules predate them. A rewrite is planned for Q3.
Illustrative examples. Times are typical, not measured; in a workshop we time your own tasks.
Which do you need?
Get people using it well
Choose, and keep choosing
Keep it safe
Not sure where to start?
Three quick questions.
Question 1 of 3
What is prompting this?
Why have a security team set up your AI?
Protecting data is our day job, so we set up your AI the way we’d secure anything else.
- Data questions firstWhat could leak is settled before switch-on.
- Set up and taught by usWe configure the tools, then train your people.
- A person always verifiesThe tool drafts. Your people sign it off.
Why keeping up matters
The #1 changed 15 times in 24 months
Today Sep 2026
Anthropic’s Claude Opus 5.5 is #1. A team choosing now would pick Claude.
Coding now: Claude Opus 5.5 in Claude Code (Anthropic), since 23 Sep.
GPT-6 Astra in Codex (OpenAI) led 5 to 23 Sep.
Also new in Sep 2026: Claude Fable 5.1, GPT-6 Sol, GPT-6 Luna, Claude Sonnet 5.5, GPT-6.1 Sol.
Best at coding:Not yet measuredAnthropicGoogleOpenAI
Short leads:Google Gemini 3 Pro, 16 to 26 Nov 2025OpenAI GPT-5.6 Sol in Codex, 12 to 16 Jul 2026Moonshot AI Kimi K3, 16 to 27 Jul 2026Alibaba Qwen3.8 Max, 1 to 2 Sep 2026OpenAI GPT-6 Astra, 5 to 23 Sep 2026
See the data
| Month | Independent tests: #1 | People’s votes: #1 | Best at coding |
|---|---|---|---|
| Oct 2024 | o1-preview OpenAI | ChatGPT-4o OpenAI, tied | Not yet measured |
| Nov 2024 | o1-preview OpenAI | Gemini Exp 1121 Google, tied | Not yet measured |
| Dec 2024 | o1-preview OpenAI | Gemini Exp 1206 Google, tied | Claude 3.5 Sonnet Anthropic |
| Jan 2025 | o1 OpenAI | Gemini 2.0 Flash Google, tied | Claude 3.5 Sonnet Anthropic |
| Feb 2025 | o3-mini OpenAI | Grok 3 (early) xAI | Claude 3.7 Sonnet Anthropic |
| Mar 2025 | o3-mini OpenAI | Gemini 2.5 Pro Google | Claude 3.7 Sonnet Anthropic |
| Apr 2025 | o4-mini OpenAI | Gemini 2.5 Pro Google | Claude 3.7 Sonnet Anthropic |
| May 2025 | o4-mini OpenAI | Gemini 2.5 Pro Google, tied | Gemini 2.5 Pro Google |
| Jun 2025 | Gemini 2.5 Pro Google | Gemini 2.5 Pro Google | Gemini 2.5 Pro Google |
| Jul 2025 | Grok 4 xAI | Gemini 2.5 Pro Google, tied | Gemini 2.5 Pro Google |
| Aug 2025 | GPT-5 OpenAI | Gemini 2.5 Pro Google, tied | GPT-5 OpenAI |
| Sep 2025 | GPT-5 OpenAI | Gemini 2.5 Pro Google, tied | GPT-5 OpenAI |
| Oct 2025 | GPT-5 Codex OpenAI, tied | Gemini 2.5 Pro Google, tied | GPT-5 OpenAI |
| Nov 2025 | Gemini 3 Pro Preview Google | Gemini 3 Pro Google, tied | Claude Opus 4.5 AnthropicGemini 3 Pro (Google) 16 to 26 Nov |
| Dec 2025 | Gemini 3 Pro Preview Google | Gemini 3 Pro Google, tied | Claude Opus 4.5 Anthropic |
| Jan 2026 | GPT-5.2 OpenAI | Gemini 3 Pro Google | Claude Opus 4.5 Anthropic |
| Feb 2026 | Gemini 3.1 Pro Preview Google | Claude Opus 4.6 Anthropic, tied | Claude Opus 4.6 Anthropic |
| Mar 2026 | Gemini 3.1 Pro Preview Google, tied | Claude Opus 4.6 Anthropic, tied | Claude Opus 4.6 Anthropic |
| Apr 2026 | GPT-5.5 OpenAI | Claude Opus 4.7 Anthropic, tied | Claude Opus 4.7 Anthropic |
| May 2026 | Claude Opus 4.8 Anthropic | Claude Opus 4.6 Anthropic, tied | Claude Opus 4.7 Anthropic |
| Jun 2026 | Claude Fable 5 Anthropic | Claude Fable 5 Anthropic, tied | Claude Fable 5 Anthropic |
| Jul 2026 | Claude Opus 5 Anthropic | Claude Fable 5 Anthropic, tied | Claude Opus 5 AnthropicGPT-5.6 Sol in Codex (OpenAI) 12 to 16 JulKimi K3 (Moonshot AI) 16 to 27 Jul |
| Aug 2026 | Claude Opus 5 Anthropic | Claude Fable 5 Anthropic, tied | Claude Opus 5 Anthropic |
| Sep 2026 | Claude Opus 5.5 Anthropic | Claude Opus 5.5 Anthropic, tied | Claude Opus 5.5 AnthropicQwen3.8 Max (Alibaba) 1 to 2 SepGPT-6 Astra (OpenAI) 5 to 23 Sep |
- A company’s best-ranked model is not always its newest.
- Independent tests: the Artificial Analysis Intelligence Index. Independent: Artificial Analysis runs every evaluation itself rather than relying on vendor-reported numbers, and publishes its methodology.
- People’s votes: the LMArena / Arena text leaderboard (style-controlled default view). Measures preference in chat, not task success; top models are usually statistically tied.
- Each line is a company’s best model at each month end, placed by its rank on that leaderboard. Ranks, not scores, because both leaderboards changed their scoring during these two years.
- Best at coding: the #1 on LMArena’s WebDev Arena, then its Code Arena from November 2025.
- Chosen for credibility and relevance, not for how often the top changes: the #1 model changed 15 times on Artificial Analysis (8 changes of company) and 11 times on LMArena (4).
Sources (59)
Leaderboards
Vendor sources
- Introducing GPT-6.1 Sol
- Introducing Claude Sonnet 5.5
- Introducing GPT-6 Sol and Luna
- Introducing Claude Opus 5.5
- Introducing Grok 4.7
- DeepSeek-V4.1-Flash: Smarter, Faster, More Efficient
- GPT-6 Astra: A new generation of intelligence
- Introducing Muse Spark 1.3
- Introducing Gemini 3.8 Flash and 3.8 Flash Cyber
- Introducing Claude Fable 5.1 and Claude Mythos 5.1
- Gemini 3.7 Flash
- Introducing Grok 4.6
- Introducing Muse Code and Muse Spark 1.2
- Alibaba Unveils Qwen3.8-Max: Its Largest and Most Capable Flagship Model to Date
- Introducing Claude Opus 5
- Introducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber
- Introducing Muse Spark 1.1
- GPT-5.6: Frontier intelligence that scales with your ambition
- Introducing Grok 4.5
- Introducing Claude Sonnet 5
- Claude Fable 5 and Claude Mythos 5
- Introducing Claude Opus 4.8
- DeepSeek V4 Preview Release
- Introducing GPT-5.5
- Introducing Claude Opus 4.7
- Introducing Muse Spark (Meta Superintelligence Labs)
- Project Glasswing: Securing critical software for the AI era
- Introducing GPT-5.4
- Gemini 3.1 Pro: A smarter model for your most complex tasks
- Grok release notes
- Claude Opus 4.6
- Introducing GPT-5.2
- Introducing Claude Opus 4.5
- Introducing Gemini 3
- Grok 4.1
- GPT-5.1: A smarter, more conversational ChatGPT
- Introducing Claude Sonnet 4.5
- Introducing GPT-5
- Claude Opus 4.1
- Introducing gpt-oss
- Grok 4
- Introducing Claude 4
- Qwen3: Think Deeper, Act Faster
- Introducing OpenAI o3 and o4-mini
- Introducing GPT-4.1 in the API
- The Llama 4 herd: the beginning of a new era of natively multimodal AI innovation
- Introducing GPT-4.5
- Claude 3.7 Sonnet and Claude Code
- Grok 3 Beta
- OpenAI o3-mini
- Change Log
- Release notes
- Introducing ChatGPT Pro
- Introducing computer use, a new Claude 3.5 Sonnet, and Claude 3.5 Haiku
How a first engagement usually goes.
Small, specific and built around work you already do.
- First callOne team, two or three tasks
The work that takes too long.
- PreparationWe check the ground
What your tools reach, and what data is safe.
- The workWorkshop, review or set-up
Your people do the tasks themselves.
- AfterwardsYou keep the guide
Request patterns, checks and next steps.