From answering questions to surfacing them

Most software waits to be asked. You open it, you navigate, you query. But in wealth management, the hardest part is often knowing what to look at next — the account that drifted, the lot worth harvesting, the plan that needs a refresh before a meeting. Proactive suggestions are the first step toward software that helps you see what matters next instead of waiting for you to find it.

The bar, though, is high. A proactive feature that fires generic prompts is just noise with good intentions. A useful one is tied to your actual context and earns the interruption.

What makes a suggestion useful

A good suggestion answers a specific set of questions before it ever reaches you:

• why now — what makes this relevant at this moment,
• what it refers to — the specific account, household, security, plan, packet, or client,
• what triggered it — the missing input or changed condition behind it,
• where it goes — the route or panel it will take you to,
• what it does — whether it drafts something, opens a review, asks a question, or requests an approval.

The difference between "you might want to review your portfolio" and "this account's drift crossed your rebalance band after yesterday's move — review the proposed trades" is the difference between noise and signal. The second is tied to a real condition and a real next step.

Built from real context, not guesses

Context-aware proactivity draws on what the system actually knows: information gaps that imply a next action, and the live screen context — which route you are on, which entities are selected, which widgets are visible, what you last did. A suggestion assembled from that context can carry a label, a prompt, the gap it came from, a route target, a prefill payload, and a relevance score.

That grounding is what keeps suggestions specific. They are generated from a concrete gap or a changed condition, not from a generic template that fires the same prompt for everyone.

The safety boundary: proactive is not autonomous

This is the line that matters most. Proactive does not mean uncontrolled. A proactive suggestion may open a route, prefill a draft, or ask a question. It must not perform an irreversible financial action without policy, entitlement, and approval support behind it.

The distinction is between surfacing and acting. Drafting a harvest plan for you to review is proactive and safe. Executing the trades because the system decided to is neither. Keeping suggestions on the surfacing side of that line is what makes proactivity trustworthy rather than alarming.

Honest about today versus tomorrow

It is worth being clear about maturity. The near-term capability is context-aware quickstarts generated from real information gaps and screen context — useful triage that points you at the right next action. The fuller vision is a genuinely event-driven, always-on capability with durable checkpoints, inbox-backed lifecycle, mute and defer controls, and real policy thresholds.

We would rather ship grounded, specific suggestions today and build toward the autonomous version deliberately than claim an always-on agent before the safety and lifecycle machinery is fully in place. Proactivity that respects the approval boundary is the only kind worth having.

The takeaway

Proactive suggestions should help you see what matters next — tied to a real account or condition, clear about why now and where they lead, and always on the safe side of the line between surfacing work and performing it. Generic prompts erode trust; context-grounded ones earn attention. Get that balance right and proactivity becomes one of the most useful things software can do: not just answering your questions, but helping you ask the right ones.