Artificial intelligence behind the platform
Every reading on your dashboard is produced by one engine reading one world: prices, volumes, events and the relationships between them. This page explains what the engine is, how it works, what it can honestly do and where its limits sit, because a tool you understand is a tool you can use well.
Registration open, limited spots
By creating an account, you agree to our Privacy Policy and our Terms of Use.
An account manager will call you within 24 hours.
1. What the engine is
The engine is a pipeline of statistical models and language-processing components built to do one job: turn a continuous flood of market data into compact summaries a human can act on. It is not a mind and not a black box with opinions; it is software that measures things consistently, at a scale no person can match, and reports what it measured in a stable vocabulary.
2. What it reads
On the numerical side: prices and volumes across the exchanges and indices covered by your plan, volatility measures, correlation between assets, and the behaviour of spreads. On the textual side: corporate announcements, index notices and macroeconomic releases, parsed into dated events. The engine connects the two, so an event reading shows not just that something happened but what the market did around it.
3. How a reading is built
A trend reading, for example, is assembled from the sequence of prices over multiple horizons, normalised for the asset's own volatility, and expressed as direction plus strength. A volatility reading places current swings in the context of the recent regime rather than against a fixed benchmark, which is why the same percentage move reads differently in a calm and a stressed market. Event readings are extracted from source documents, timestamped and linked to observed price behaviour.
The pipeline runs in stages: ingestion and cleaning, where bad prints and duplicates are removed before anything else sees them; feature construction, where the cleaned series becomes the measured quantities behind readings; summarisation, where those quantities become the sentences on your dashboard; and logging, where every stage's output is stored so any reading can be reconstructed exactly as it was issued.
Every reading links back to its inputs. If a trend reading changed, you can see which sessions moved it. This traceability is the difference between an analysis you use and an oracle you merely obey.
4. What the engine is good at
Scale and consistency. It watches every asset on your list every hour without fatigue, applies the same definitions on a boring Tuesday and a dramatic Friday, never forgets a log entry and never rewrites its own history. For a private investor, that consistency is the rarest resource in the market: most human errors are failures of attention and discipline, and those are precisely the failures the engine does not have.
Comparison is a second strength. Asked whether this asset's volatility is unusual, a person reaches for a feeling; the engine reaches for the asset's own history and answers with the distribution. Asked how two holdings behaved together through the last three rate decisions, it answers the same way every time, which makes its answers usable as a baseline for your own judgement.
Patience with tedium is the third. Reconciling an event list against three sources at midnight is work humans do badly and resent; the engine does it identically every night, which is why the morning summary is reliable in a way a tired person's recollection is not.
5. What the engine cannot do
It cannot know the future. Every reading summarises what has already happened; the future is generated by participants who have not acted yet. It cannot know your circumstances, your tax position, your obligations or your temperament, which is why it never tells you to buy or sell. And it cannot remove risk: markets move against well-informed positions all the time, a fact this page shares with the risk disclosure deliberately.
It also cannot feel the market's mood before it shows in data. By the time unusual behaviour appears in prices and volumes, whatever caused it is partly priced in; the engine is fast at organising evidence, never early at knowing news. Anyone who promises you an engine that knows first is selling a story with a software label on it.
6. Why we say "assisted" analysis
The word assisted carries the division of labour. The engine gathers, measures and organises; you supply intent, judgement and the decision. Used that way, the technology amplifies an investor's own process. Used as a substitute for judgement, it merely automates confusion at higher speed, which is why our account managers spend the first weeks teaching the reading, not selling the magic.
7. How readings stay honest
Definitions are fixed and published: what counts as a trend change, how volatility bands are drawn, which events make the list. The engine does not silently retune itself to flatter its own history, and its readings are never edited after the fact. When the underlying data was delayed or incomplete, the reading says so, because a confident number built on a gap is worse than an admitted gap.
Changes to the engine are versioned like the audit history: when a definition genuinely improves, the change is announced, dated, and applied forward, never back-filled into old readings to make the new logic look prescient. A summary whose past quietly changes is not a record; it is marketing.
Language is policed the same way. Readings describe what happened and attach confidence in plain terms; they avoid prediction vocabulary entirely, because a phrase like "expected to continue" is a forecast wearing an analysis costume.
8. The engine and your decisions
Your decision log and the engine's readings sit side by side in the weekly summary. Over months, that pairing produces something no advertisement can: your own track record of judgement against organised information. Investors who read it learn which readings they use well and which they habitually overreact to, which is genuinely useful self-knowledge, of the kind that compounds.
The summary is deliberately unsentimental. It does not grade your choices against the best possible outcome, which nobody achieves, but against the information available when you chose, which is the only fair standard and the only one that improves future decisions.
9. Costs and access
The engine powers every plan; there is no separate AI tier. Basic at £19 per month from a £250 deposit includes the full reading set for core markets, Advanced at £45 from £1,500 extends the universe, Premium at £95 from £5,000 covers everything the engine reads. Detail is on the fees page.
10. Questions people ask about the engine
Is it a trading bot? No. It produces analysis; it holds nothing, trades nothing and executes nothing.
Does it learn from my account? It personalises what it shows you, your watchlists and alerts; it does not train itself on your balance or your decisions.
Can it be wrong? Yes, and so can any analysis. The traceability behind each reading exists so you can check its reasoning rather than trust a label.
Is it a large language model? The engine combines statistical models for numerical series with language-processing components for documents. It is not a chatbot and it does not compose opinions; every sentence on the dashboard is assembled from measured quantities with fixed wording.
What happens during a data outage? Readings that cannot be built honestly are marked as stale rather than extrapolated, and the gap closes when the feed does. You will always know which reading was built on what.
To see the engine's output in context, the share analysis and crypto monitoring pages show the same readings applied to each asset class.