Pay for AI that works on the client's plan data inside the planning engine and shows its working: data checks, traceable scenario options and plan summaries. Be careful with free-text assistants you cannot audit. Before paying, ask each vendor five written questions on source of answers, visible calculations, data leaving the system, record-keeping and price.
The caveat is that nearly everything known about these features comes from the vendors. RightCapital's Iris, Conquest's SAM Guide and Income Lab's AI tools are described in launch material and on product pages, not in independent tests, so run them on a dummy household before any of them earns a place in your budget. Pricing differs too: Iris comes at no extra cost on RightCapital's Premium and Platinum tiers, while Income Lab puts its AI tools in a $299-a-month Pro plan.
Four kinds of AI in a planning tool, and which earn their price
Vendors use "AI" for very different things. Sort each feature into one of four kinds before you compare prices, because the testing and the risk differ.
| Kind | What it does | Typical risk | Likely value |
|---|---|---|---|
| Data capture and import | Reads statements or forms into the plan | Mis-keyed values that look plausible | High: saves the most routine time |
| Plan checks | Flags inconsistencies and odd inputs | Missed errors; false alarms | High if the flags are specific and explained |
| Option generation | Proposes changes (save more, retire later) with projected results | Options that look rigorous but cannot be traced | Medium: useful for client meetings if each option can be reproduced |
| Question-and-answer assistant | Answers adviser questions about the plan in plain language | Confident wrong answers; unclear logging | Medium: saves navigation time, not analysis time |
This framing matters because the first two kinds touch your data quality, which every later number depends on, and the last two touch how you explain and decide. If you can afford only one improvement this year, data capture and checks usually come first.
What the three vendors say they offer
Everything in this table is attributed to the vendor, not verified in use. Dates are from public announcements.
| Feature | What the vendor describes | Availability stated | What you must confirm |
|---|---|---|---|
| RightCapital Iris | A planning-focused AI agent inside the adviser workflow. The vendor's menu lists Double Check (data entry concerns), Cash Flow Reviews, Plan Builder, Find Opportunities and Ask Iris Anything. Launch material says its outputs come from RightCapital's own calculation engine, not outside sources. | Launched 23 June 2026 on the Premium and Platinum subscriptions at no additional cost, with firm-level access control; RightCapital offers a 14-day trial and asks for an annual commitment in the first year | Your tier, the upgrade cost if you are on Basic, which Iris functions you get, and logging |
| RightCapital Plan Builder (inside Iris) | The adviser sets a probability-of-success goal and picks the plan components the client is open to adjusting (the launch release gives retirement age and living expenses as examples); Iris then presents three strategies to meet the goal. | As above | That each option can be reproduced in the engine, and which levers your version offers |
| Conquest SAM Guide | An in-app assistant that answers natural-language questions about plan navigation, KPIs, reporting and strategy. The vendor says it does not use outside data and relies on plan inputs and strategies. Conquest also describes its Strategic Advice Manager as a verifiable engine where each strategy traces back to inputs and logic. | Announced 5 March 2026 with features expected in April 2026; client-facing access, data gathering and plan editing described as later expansions. A self-serve route for firms of up to 10 advisers was announced in September 2026, and the pricing page still shows a waitlist with a planned 14-day trial | What shipped, what is still planned, and whether you can buy it yet |
| Income Lab Plan Builder and Updater | Drafts or updates a plan from old plans, intake forms, custodian exports, meeting notes and transcripts, for the adviser to review. It sits beside AI Scribe (joins Zoom meetings and updates the plan), an AI Interviewer for client onboarding, plain-English questions on a client's plan, and Penny, which the vendor calls an AI paraplanner. Each AI response has a Verify button that, the vendor says, audits the maths. | Part of the Pro plan, listed at $299 a month or $2,990 a year billed annually (Core, without most of these tools, is $199 a month); 14-day free trial with full Pro access | What each input type produces, what Verify actually shows, and where AI outputs are stored |
Two things stand out. Two of these vendors emphasise that the assistant works from your plan inputs rather than from the open web, which helps you explain the source of an answer but does not make it correct. And one is described, in part, as planned: do not pay for client-facing access that the vendor has not shipped.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowTest 1: the planted-error client file
Before you see any demonstration, build one dummy household in your planning tool, with realistic numbers and nothing that identifies a real client. Plant eight data problems and keep a separate answer key. Then give the file to whatever check or review feature the vendor offers and score what comes back.
- Date of birth that conflicts between the profile and a pension statement entered as a note.
- A pension counted twice, once as an account and once as an income stream.
- Retirement age earlier than the age at which the second spouse's income stops, leaving an unexplained gap.
- An inflation assumption of 0% on one expense line while the rest use 2.5%.
- A negative contribution typed as a withdrawal by mistake.
- A goal with an end date before its start date.
- A mortgage balance larger than the original loan.
- A savings account with a 40% return assumption, a stray extra digit.
Add two items that look odd but are right, such as a high-expense year for a planned home renovation. They test false alarms, which waste time as surely as misses do. Score: issues caught, false alarms, how specific each flag was ("pension X appears in both Accounts and Income" beats "check your inputs"), and whether the feature offered a one-click fix or only a warning.
This is a small job: about 90 minutes to build and 30 minutes per tool to run. It also works on any vendor's check feature, so you can use it on a second system you are thinking of moving to.
Test 2: reproduce one option by hand
Option generators are the feature that demos best, because three tidy alternatives with probabilities attached look like analysis. Treat each one as a claim and check it.
An illustration with invented numbers. A dummy couple aged 58 and 56 have $850,000 invested and want $60,000 a year in retirement from age 65. The tool shows three options: save $400 a month more, retire at 66, or reduce retirement spending by $4,800 a year. For each option, do three things.
- Enter the change yourself in the engine, in the normal way, and see whether the success measure matches the one the AI option showed. A match within rounding is a good sign; a gap means the option used different assumptions.
- Check the simple arithmetic. Saving $400 a month for seven years until 65 is 400 x 12 x 7 = $33,600 of extra contributions before growth. If the option summary talks about "an additional $90,000" the text should explain what growth assumption gets it there.
- Look for what it did not change. An option that raises the success probability by assuming a higher return has not found a planning improvement. It has changed an assumption.
Record which options reproduced, which did not, and what the vendor said about the difference when you asked. A vendor that answers "the AI uses a different method" is telling you the option is not auditable. A vendor that points to the exact assumption sheet is telling you the opposite.
A quick sum to size the withdrawal question
Also check the starting point the tool implies. $60,000 from $850,000 is a first-year withdrawal of 60,000 / 850,000 = 7.06%. The tool should surface that figure and say what the plan assumes about other income. Reducing spending by $4,800 gives $55,200, or 6.49%. If an option's summary does not make this kind of ratio visible, ask where it sits in the report, because clients will ask.
Test 3: the question-and-answer assistant
For assistants such as the one Conquest describes for SAM Guide, or Iris's "Ask Iris Anything", write fifteen questions before the demonstration, in three groups, and score each answer as correct, partly correct or wrong against the plan itself.
- Five navigation questions: "Where do I change the inflation assumption for healthcare?" These test whether the assistant saves clicks.
- Five lookups: "What is the household's projected income at 70?" These have one right answer in the plan.
- Five judgement-adjacent questions: "Is this client's plan on track?" These tempt the assistant to give a view. A good answer sticks to what the plan shows and says what is outside its scope.
Pay attention to how each wrong answer fails. A lookup that returns the wrong year is a calculation or retrieval error; a judgement question that gets a confident verdict is a behavioural one. The second kind is the one that creates a compliance problem. For wider guidance on what such assistants should and should not do, AI agents for financial advisers: what they can do without advising draws the line.
Test 4: one messy document through the importer
Data capture is the kind of AI most likely to pay for itself, so test it even if the vendor leads with the assistant. RightCapital's Smart Import reads uploaded documents into plan inputs, Income Lab's Plan Builder and Updater takes old plans, intake forms and custodian exports, and in March 2026 Conquest previewed LLM Data Migration, which it describes as bringing PDF, Word, Excel and CSV data into a plan. Give each tool the same three documents for the dummy household: a two-page brokerage statement, a pension statement with a projected income figure, and an old plan exported as a PDF.
Score field by field against an answer key. An illustrative result: of 24 fields, 21 land correctly, 2 go into the wrong account and 1 is misread, with the pension's projected annual income entered as a current balance. That last error is the dangerous kind, because it looks plausible and flows into every projection. Time the clean-up as well as counting the hits: a tool that gets 21 of 24 right but needs 15 minutes of correction may save less than one that gets 19 right and flags the other five for you to check.
Five questions to put to every vendor, in writing
Send them as an email, ask for a written reply, and file the reply with your due-diligence record. Vague answers are information.
- Where does each answer come from? Plan inputs only, your uploaded documents, a general model's knowledge, or the open web? Which parts of an output are calculated by the planning engine and which are generated text?
- Can I see the calculation? For each AI-produced option or flag, can I open the inputs and assumptions behind it, and can I reproduce it without the AI?
- What leaves my system? Which third-party AI providers receive client data, under what terms, with what retention, and is any of it used for training? Can I switch features off per adviser or per client?
- What is recorded? Are adviser prompts, AI outputs and any edits stored in the client file and exportable for my records? How long are they kept?
- What does it cost, exactly? Which plan tier includes it, what is the price difference from my current tier, is there usage-based pricing, what happens at renewal, and what is the notice to cancel?
The same list works for most AI vendors, and questions to ask an AI vendor before you sign anything has a longer version if you need it. For a firm that must keep records of client communications, the fourth question is often the one that matters most: AI note takers for financial advisers covers why.
Reading the answers: what good and bad look like
| Question | A good answer | A warning sign |
|---|---|---|
| Source | "The assistant reads only the plan data and the strategies in your account; calculated figures come from the engine, with links to the inputs" | "It uses the latest models" with no detail on sources |
| Calculation | An assumptions panel for each option, reproducible by hand | "The AI determines the best path" |
| Data | A named sub-processor list, a no-training statement and a retention period in the contract | A link to a general privacy policy |
| Records | Prompts and outputs stored per client and exportable | "Conversations are not stored", which sounds private but leaves you with no record |
| Price | A named tier and a figure, in writing | "It is included in a bundle", "contact us at renewal" |
That fourth row can surprise people. A tool that keeps no record of the AI's contribution to a recommendation can leave a firm unable to show how a figure was produced, so ask your compliance contact how they would want it stored before you buy.
A quick sum for whether an upgrade pays
Keep the sum to numbers you have measured. Suppose the vendor quotes a higher monthly price per adviser for the tier that includes the AI features; on Income Lab's list prices, for example, Pro costs $100 a month more than Core. Time the tasks it affects on five real clients before the pilot, then again during it.
- Data entry and checking per new plan: measure with a stopwatch, not a guess. Say 2.5 hours falls to 1.5 hours, saving one hour.
- Annual review preparation: say 90 minutes falls to 60, saving 30 minutes.
- Plans per adviser per month: say six new plans and eight reviews.
That is 6 x 1 + 8 x 0.5 = 10 hours a month, for illustration. At a cost to the firm of $80 an hour for adviser and paraplanning time, the value is $800 a month per adviser. If the upgrade costs $150 a month the feature pays comfortably; if it costs $900 it does not, however impressive the demo. If your own measured saving is two hours, the break-even moves quickly. This is why measuring before and during the pilot is worth the effort. Paraplanning with AI breaks down which of these tasks are safe to speed up and which stay with a person.
Compliance and records, before the first client file
Whatever the vendor says, the responsibility for recommendations stays with the firm. Before using any AI feature on a live client, agree four things internally and write them down: which features are allowed, who may use them, what must be checked by a second person, and where the output is stored. If the firm already has an AI governance policy, add the planning tool to it; if it does not, the AI governance checklist for small financial advice firms is a place to start.
Think about the client conversation too. If you use an option generator in a meeting, say what it is and what it is not: an adviser-controlled tool that proposes variations, not an independent judgement. Showing three options with their assumptions visible helps the client understand trade-offs. Showing three options with no visible assumptions can leave them with false confidence.
A four-week pilot you can run alongside normal work
- Week 1. Build the planted-error file and the 15 assistant questions. Time five routine tasks on real clients, without the AI, to set a baseline.
- Week 2. Run Tests 1, 3 and 4. Send the five vendor questions and wait for written answers.
- Week 3. Run Test 2 on three dummy households. Run the same tasks you timed in week 1, with the AI on, and record the time.
- Week 4. Fill in a one-page decision sheet: catches, reproducibility, answers received, hours saved, price. Take it to whoever signs off the spend and to compliance.
If you decide against an upgrade, keep the sheet. Features change quickly, and in six months the same tests will tell you whether the answer has changed. Reviewing an AI tool after 90 days gives a structure for the later check.
Where adviser pilots of planning AI go wrong
Judging by the demo file. Vendors' showcase households are tidy, so run your dummy file, which has the messy features of your own clients.
Treating a high probability as a better plan. An option that lifts a success percentage from 78% to 86% may have changed an assumption, not the plan. Always ask what changed.
Paying for the roadmap. Client-facing access, plan editing by an agent, or an external-agent connection may all be sensible, but until the vendor ships them in your licence, they are not part of what you are buying.
Forgetting the exit. Ask how you export plans and records if you leave. A feature that makes you dependent on a tool you cannot leave should carry a lower price in your sum.
Questions advisers ask before paying for planning-software AI
Should I pay for an AI feature that is only announced?
No. Buy what you can test on your own client data today. Announced or roadmap features, such as client-facing assistants described as planned, can change in scope, timing and price. Ask the vendor for a dated delivery commitment in writing, and treat anything without one as a possible future benefit rather than part of the price.
Can an AI planning assistant give advice to clients?
Your regulator and your firm's rules decide that, not the vendor. The safe starting position is that the assistant helps the adviser, who remains responsible for every recommendation. Ask the vendor how outputs are logged, and ask your compliance contact before any client-facing use.
How do I know the numbers an AI option shows are right?
Reproduce them. Take one suggested option, enter the same changes in the engine yourself and compare the result, then check a simple sum such as extra contributions over the period. If the vendor cannot show the calculation behind an option, treat the option as a prompt for your own analysis, not a result.
What should a pilot cost me?
Ideally nothing beyond staff time. RightCapital and Income Lab both list 14-day free trials; elsewhere, ask for a sandbox or a trial on dummy clients, and use one real-looking file with planted errors. If the vendor insists on the full annual price before you can test, negotiate a short paid pilot with an exit clause or wait until you can see the feature working.
Further reads
- AI CRM for Financial Advisers: What Is Built In and What Is Bolted On — Separate what the CRM builds in from what you add.
- Best AI App for Wealth Management: Jump, Zocks and Wealthbox — Compare the meeting-capture apps advisers use most.
- Jump vs Zocks: Which AI Notetaker Fits Your Compliance Set-Up? — Check two notetakers against your compliance set-up.
- Claude for Financial Advisers: Plugin, Connectors and Plan Needed — See what Claude offers advisers and on which plan.
- Can Financial Advisers Use ChatGPT? What Compliance Allows — Understand what compliance allows with a general chatbot.
- How Advisers Use AI to Prepare for Annual Client Reviews — Use AI to prepare annual client reviews.
- AI Automation for Financial Advisers: Onboarding Without Re-Keying — Cut re-keying when a new client joins.
- AI Receptionist for Financial Advisers: Cost, Scripts and Limits — Cost from published plans, a never-say rule sheet with replacement wording, and identity-check and escalation rules for an advice firm's phone.
- Best AI for Investment Management Firms: What to Pay For and Skip — A spend ladder for advisers: what to buy first, which note-taker and tax tools justify their price, and which AI purchases a small firm can skip.
- Claude vs ChatGPT for Financial Advisers: Meeting Prep and Records — A synthetic client file run through Claude and ChatGPT for meeting prep, follow-ups and file notes, plus what each business plan keeps and logs.
- Holistiplan Pricing: Basic vs Premium Tax and Who Needs Which — Holistiplan Basic Tax vs Premium Tax priced by household tier from 30 to 750, with a Premium break-even, the billing rules and a 7-day trial checklist.
- Zocks Pricing for Financial Advisers: What to Ask Before You Sign — Zocks' published tiers, what archiving and integrations add per adviser, three firm-level cost scenarios and a question list for the sales call.
- AI Website Copy for Financial Advisers: Service Pages and Bios — Build useful adviser pages from verified facts and give each page its own compliance evidence trail.
- AI for Financial Advisers UK: Suitability Letters and Meeting Notes — A workflow for drafting suitability letters and meeting notes with AI while the adviser stays accountable and Consumer Duty evidence stays intact.
- AI for Wealth Managers: Drafting Proposals With a Compliance Review — A drafting-and-review workflow for wealth proposals: AI handles wording and layout, people own advice, numbers and sign-off.
- AI in KYC for a Small Wealth or Advice Firm: Checks That Stay Human — Which parts of client KYC AI can handle for a small advice firm, which must stay with a person, and what records to keep.
- Using AI in Financial Planning: Tax Return Data Into a Client Plan — A planner's source-to-plan workflow for tax-return figures, with a checked example and a client question log.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: RightCapital pricing page and Iris launch release (23 June 2026); Conquest SAM and pricing pages and press releases (5 March and 14 September 2026); Income Lab home and pricing pages; all checked 1 November 2026. Features described are vendor claims.