24/5 Markets Still Need a Source of Truth

On 21 July 2026, the London Stock Exchange announced plans for London Stock Exchange 24, a separate 24/5 venue intended to support digital, algorithmic and agentic trading. Client testing is planned by the end of 2026, with exchange-traded products expected to be the first asset class in the first half of 2027, subject to regulatory approval.

The announcement is about more than longer trading hours. It points towards a market in which software agents may interact with market data, order-management systems and execution workflows while regulated infrastructure remains responsible for governance, resilience and control.

That creates an important question beyond the trading venue itself:

What must an agent know about an asset before a person or system can safely rely on its output?

Key point: Faster and more automated markets do not reduce the need for trusted asset information. They increase the need for source-linked evidence, clear permissions, current versions, visible review status and accountable human decisions.

The LSE 24 announcement is a useful market signal. It is not evidence that every asset, platform or AI workflow is ready for autonomous trading.

A market can be available while an asset is not ready

Extended trading hours solve an availability problem. They allow eligible market participants to respond across more time zones and outside the traditional trading day.

They do not automatically solve an asset-information problem.

An asset may still have:

• conflicting descriptions;
• missing or unsigned documents;
• an unclear ownership chain;
• transfer restrictions;
• outdated registry information;
• disputed amounts or performance history;
• confidential material mixed with public information;
• AI-generated findings that no person has reviewed; or
• a status that has changed since the data was collected.

A venue may be ready to receive an instruction before the underlying evidence is ready to support a decision.

This distinction is especially important for private, document-backed and non-standard assets. Their identity, rights, restrictions and current status may be distributed across contracts, amendments, notices, registers, payment records, correspondence and professional opinions rather than one authoritative data feed.

The speed of the execution layer cannot make those uncertainties disappear.

Data is not the same as evidence

An agent can read a field stating that a claim is worth EUR 500,000. That does not tell the agent:

• whether EUR 500,000 is the original amount, current balance, claimed amount or valuation;
• which document supplied the number;
• when the figure was last updated;
• whether payments, interest, fees or disputes were considered;
• whether a user entered the amount or AI inferred it;
• whether anyone reviewed the result; or
• whether the viewer is permitted to see the supporting documents.

The field may be useful, but it is not self-authenticating.

A reliable workflow needs to preserve the relationship between a statement and its basis. Depending on the asset, that basis may include a source document, official registry entry, calculation, signed confirmation, third-party report or recorded user assertion.

The system should also show the limitations of that basis. A document can support what was written or agreed without proving that the document is authentic, enforceable, current or complete. A registry record can be authoritative for a defined purpose without containing every private restriction or agreement.

This is why machine-readable should not be treated as a synonym for verified.

What an agent-readable asset record should answer

Before an AI agent uses asset information in analysis, screening, recommendation or workflow preparation, the record should make several questions answerable.

Without this context, an agent may process information quickly while misunderstanding its authority, scope or age.

A source of truth is a governed chain, not a magic database

The phrase source of truth can sound as if one database contains the final answer to every question. Real assets rarely work that way.

Different sources can be authoritative for different purposes:

• an official register may show a registered proprietor or filing status;
• a signed contract may define the parties' rights and obligations;
• an amendment may replace part of the original agreement;
• a payment ledger may show later performance;
• a court order may change enforcement rights;
• a qualified professional may provide a limited opinion; and
• an asset owner may supply information that still requires confirmation.

The useful source of truth is therefore a governed record that shows how these sources relate, which one supports each statement and where conflicts remain.

It should not flatten every input into one apparently certain answer.

For example, if an official register and an uploaded assignment appear inconsistent, the record should expose the inconsistency and the dates involved. An agent should not silently select the more convenient source and present the result as verified ownership.

Provenance must travel with the data

Provenance explains where information came from and what happened to it.

For an asset fact, useful provenance may include:

• the source document and version;
• the relevant page, clause or extracted passage;
• the date on which the source was obtained;
• the person or system that created the field;
• the AI model or process that produced a suggestion;
• the person who corrected or confirmed it;
• the scope and date of any third-party review; and
• later events that changed or superseded the information.

This distinction matters because the following statements do not have the same evidential weight:

• uploaded by the asset owner;
• extracted by AI from an uploaded agreement;
• confirmed by an authorised user;
• checked against an official source on a stated date; and
• reviewed by an independent professional within a defined scope.

An agent should be able to distinguish them. A human reviewer should be able to see the same distinction.

See What Can Third-Party Verification—and What Can It Not—Confirm?.

Review status must remain visible

AI can help classify documents, extract facts, identify inconsistencies and draft a structured asset description. These outputs can reduce manual work, but they remain outputs of a process—not final legal or commercial conclusions.

A responsible record should distinguish at least:

• a user-supplied statement;
• a source-extracted fact;
• an AI-generated suggestion;
• a user-reviewed and confirmed fact;
• an independently reviewed item;
• a disputed item;
• a superseded item; and
• an unresolved gap.

If these states are hidden, downstream automation may treat an unreviewed suggestion as equal to a confirmed fact.

The same principle applies to an entire asset package. Draft, reviewed, finalised, signed, registered and tradable are not interchangeable words. Each status should have a defined workflow meaning, visible evidence and clear limitations.

See What Finalising an Asset Package Means—and What It Does Not.

Permissions are part of the asset data model

An agent should not receive information merely because the information exists.

Asset packages may contain personal data, trade secrets, privileged communications, security information, source code, financial records and commercially sensitive agreements. Some information may be suitable for public discovery. Other material may be available only to an owner, an identified reviewer or a recipient acting for a defined purpose under an NDA.

A machine-readable record should therefore carry machine-enforceable access context, including:

• which person, organisation or role may access the information;
• which asset and document set the permission covers;
• whether access is public, private, selected or NDA-controlled;
• the permitted purpose;
• any expiry, withdrawal or revocation state;
• whether download or onward disclosure is restricted; and
• which access and disclosure events were recorded.

An NDA can support confidentiality, but it does not prove ownership, create authority to disclose third-party information or provide a lawful basis for every use of personal data.

Access control must also be enforced by the platform. It should not depend on the AI model remembering an instruction from an earlier prompt.

See Public, Private or NDA-Controlled: Choosing How to Share Information.

Human accountability becomes more important, not less

Agentic workflows can help a person move from a question to a prepared action. The important boundary is between assistance and authority.

An AI assistant may be able to:

• summarise an asset package;
• find source material;
• compare versions;
• identify missing evidence;
• draft a review checklist;
• prepare a structured asset record;
• suggest a category or description; or
• prepare an action for an authorised user to review.

It should not be assumed to have authority to:

• confirm legal ownership;
• approve an asset for sale or financing;
• disclose restricted documents;
• sign an NDA or transaction agreement;
• register or list an asset;
• make an offer;
• transfer an asset or funds; or
• execute or settle a transaction.

Material legal, financial, disclosure and lifecycle actions should remain behind explicit permissions, eligibility checks and user confirmation. The system should record what the agent proposed, what the user saw, what the user authorised and which service performed the action.

Automation can shorten a workflow. It should not erase accountability.

The Asset Passport as an evidence layer

An Asset Passport can help bridge the gap between documents intended for human reading and structured information needed by digital workflows.

Its useful role is not to declare that every asset is valid, verified or market-ready. It is to present the available record in a form that makes its sources and limitations easier to inspect.

A strong Asset Passport may bring together:

• the asset's identity and description;
• source-linked facts;
• supporting documents;
• ownership and authority information supplied for review;
• recorded lifecycle events;
• AI-assisted findings;
• human review status;
• missing items and contradictions;
• relevant official or third-party references; and
• controlled access to the underlying evidence.

For an agent, this can provide a more dependable input than an isolated spreadsheet row or generated summary. For a person, it can show why the system reached a conclusion and where further review is required.

The same record can support both audiences only if the human-readable explanation and machine-readable structure remain aligned.

What DaDepo supports today

DaDepo currently focuses on preparing and reviewing document-backed asset packages rather than autonomous trading.

Depending on the enabled workflow, users can:

• upload and organise source documents;
• use AI-assisted document analysis and asset-draft suggestions;
• review and correct suggested information;
• create a structured asset record and Asset Passport;
• distinguish available source-backed facts, provenance and review status;
• identify missing or inconsistent information;
• keep information private or share it through controlled workflows;
• use NDA-controlled access where applicable; and
• retain available version, lifecycle, access and audit records.

These functions can make an asset easier for a person or future digital service to assess. They do not mean that DaDepo has independently verified the asset, admitted it to a regulated market or authorised an agent to transact.

The user remains responsible for reviewing information and deciding whether it is accurate, lawful to disclose and suitable for the intended next step.

What remains a future direction

DaDepo does not currently claim to provide a live connection to LSE 24, autonomous order execution or a general-purpose trading agent.

Future agent-ready capabilities may include:

• permission-scoped APIs for reading Asset Passport information;
• machine-readable fact, source, confidence and review states;
• version-pinned data packages for a particular decision or transaction stage;
• controlled agent tools that can prepare, but not silently complete, material actions;
• explicit confirmation workflows;
• stronger audit records for agent runs and tool calls;
• registry, verification, identity and market-infrastructure integrations; and
• policies defining which agents may use which information for which purpose.

These are possible directions, not statements that every integration or capability is currently available.

Any regulated trading, custody, clearing, settlement, official registration or depository service would require the relevant legal basis, permissions, providers and operating rules. LSEG's planned Digital Securities Depository is market infrastructure being developed by LSEG and is distinct from DaDepo's asset-information and preparation role.

A practical example

Consider an agent asked to assess whether a document-backed asset appears ready for buyer review.

The asset record contains a contract amount, an assignment document, several payment records and a generated summary.

A weak workflow might read the amount, scan the summary and return ready.

A stronger workflow would ask:

1. Which signed document creates the relevant right?
2. Does the assignment cover the same asset, parties and jurisdiction?
3. Is the recorded amount current as of a stated date?
4. Do the payment records reduce the outstanding balance?
5. Are any amendments, disputes, restrictions or consents missing?
6. Which findings came from AI and which were confirmed by a person?
7. Are official records relevant, and when were they last checked?
8. Which documents may the prospective reviewer access?
9. Is an NDA or another permission required?
10. Is the agent allowed only to report readiness, or also to prepare a disclosure request?

The result may be ready for initial review, conditionally ready, or not ready because specified evidence is missing. That qualified result is more useful than a confident answer that hides uncertainty.

It also gives the next person a practical route forward.

A readiness checklist for agent-enabled workflows

Before asset information is exposed to an AI agent or another automated service, ask:

• Is the asset identified consistently across the package?
• Can important facts be traced to named sources?
• Are source dates and versions visible?
• Are AI suggestions distinguished from reviewed facts?
• Are conflicts and missing evidence preserved rather than silently resolved?
• Is the official source identified where registration or status depends on one?
• Is the information current enough for the intended decision?
• Are public, private and NDA-controlled materials separated?
• Is the agent's permission and purpose limited?
• Do material actions require explicit human confirmation?
• Will the system record the input version, proposed action and final decision?
• Can an authorised reviewer understand why the agent reached its result?

If the answer to these questions is unclear, adding a faster agent may accelerate uncertainty rather than reduce it.

Trust must move with the market

Near-continuous venues, digital securities infrastructure and agent-enabled workflows may change how quickly market participants can observe, analyse and act.

But an automated workflow is only as dependable as the authority, evidence, permissions and controls attached to its inputs.

DaDepo's opportunity is not to claim that every asset should immediately become autonomously tradable. It is to help build the structured evidence layer that can make digital review more understandable, controlled and accountable.

The market can move faster only if trust can move with it.

Prepare the evidence before automating the decision

Start with the documents and information already available. Organise the asset, connect important facts to their sources, identify missing evidence, separate public from restricted material and make review status visible.

That work is valuable before any trading integration exists. It gives people a clearer package today and creates a stronger foundation for controlled digital and agent-enabled workflows tomorrow.

DaDepo provides technology and information tools. It does not provide legal, financial, investment, tax, accounting or valuation advice, and it does not guarantee ownership, legal validity, verification, market admission, transferability, value, sale, settlement or liquidity.