I Asked AI When to Claim Social Security

Today I asked AI a very important retirement question. When should this married couple who’s about to retire claim social security? It’s one of the biggest questions that you have and it’s one of the biggest questions that we receive.

When I received the answer, it was very insightful. It hit on a lot of very key points. But when we ran the same case through two of our planning softwares, a bigger issue emerged. The question wasn’t when should they take social security, it’s does this retirement plan work at all?

The Case Study: Assets, spending goals, and longevity

So in the last video we put out about retirement planning and whether AI could put together a retirement plan, we had one comment that said, AI usually does better when you tell it that it’s a retirement planning expert. So that’s what I did here.

So here’s the problem. You are a retirement planning and social security expert. A married couple is trying to decide when to claim social security. Husband is 64 and retiring now. Wife is 62 and retired. His social security at full retirement age is $2,850 a month. Hers at full retirement age is $1,200 a month. They have $850,000 in traditional IRAs, $250,000 in Roth IRAs, $250K in a joint brokerage account, and $100K in cash savings.

No pension, home is completely paid off. They estimate they need $8,500 per month after tax on average to support retirement spending. They expect higher spending in the first 10 years to travel and provide family support. Both are in good health, have family longevity. She expects to live until 95 and he 90. Their goals are to maximize lifetime income, pay less tax, protect the surviving spouse, and never be a burden to children. Should they claim social security early, at full retirement age, or delay? Please recommend the best strategy and explain reasoning.

AI’s recommended strategy: Bridge years and Social Security

So I actually asked AI to give me this particular prompt, and this is what it came up with. And it’s a very good prompt in the sense that this is something that we would see quite often. The goals, maximize income, pay less tax, not be a burden to children, very common. And the questions always are, should I take it early? Should I delay? What should we do? So we’re going to run through this and see exactly what comes up.

OK, so here’s the recommended strategy. Ages 62 to 70 is the bridge years, should take from the brokerage account, traditional IRA withdrawal strategically, then do Roth conversions. Okay, so little bit interesting here because if we were trying to maximize the overall tax plan, one, we have to take into account how much money is in the brokerage account and the cash savings, which was about 350,000.

But we have to do two things here. We have to bridge income until age 70. So this is six full years. So we need to really look more closely at how much are we doing conversions for? How much do need to pull to pay those taxes on the conversions? But also taking into consideration if we’re doing what we call a multi-account distribution strategy here, where we’re taking from the brokerage and the IRA, how much are we taking from the IRA, how much are we converting, and of course, what’s the impact over long periods of time.

In general, this makes a lot of sense. In reality, we actually need a lot more specificity here and a lot of modeling. Okay, so wife claims at 62 is the consideration or full retirement age depending on tax plan. Phase two, husband claims at age 70, maximum benefit. Okay? withdrawal strategy, spend from brokerage and IRA, do Roth conversions. We just, just basically it’s repeating what it’s already said.

Okay, bottom line recommendation, why flexible? Same thing, just a summary. Okay. So now it’s giving me an option. I can run a detailed year by year income and tax projection comparing early versus full retirement age, lifetime taxes, portfolio longevity. Okay, let’s just say yes. Yes, I want to know more specifically what I should do.

The actionable plan and the “Golden Window”

Okay, so here is our clear actionable plan with timing, not just theory. So recommended strategy, husband’s age 64 now, delay until 70. Wife best claiming age 62 or 63, so soon, so basically turn it on now. Here would be the benefit $840 a month, which is about 70 % of the full retirement age benefit. Okay, so now an income timeline. What should happen over the next six to eight years depending on if we’re looking at the husband or the wife basically until age 70.

So the wife would have about $840 a month or $10,000 a year in social security. We would need about $90,000 a year to reach the spending goal. 90 plus 10 is 100, which was about $8,500 a month, the original spending goal, partial Roth conversions.

So this is what they call the golden window. age 62 to 70, we’re converting moving money from that tax-infested IRA over to the tax-free Roth. That makes a lot of sense. Age 70, husband turns on the social security. Then the portfolio withdrawals drop significantly because so much is coming from social security. It’s about 42,000 a year for his, 10,000 for hers. So that’s about 52,000 a year for social security. This is the longevity insurance component after the first death.

The husband in this example, the survivor keeps the 35.30 per month, but it’s inflation adjusted. So it would actually be a lot more than that. And then, okay, so on the Roth conversion strategy, it’s telling us to fill out the 12 or the 22 % brackets, which is somewhere between $40,000 to $100,000 a year.

Evaluating the trade-offs of early spending

Why this matters, age 62 to 70, the withdrawal strategy, so the sequencing, take from the Berkowitz account, manage capital gains, this is an important item here, traditional IRA, paired with the Roth conversions. Minimal Roth withdrawals. So let the Roth IRA continue to grow and defer.

Okay. So there’s still a lot of, there’s a lot of specificity here or really analysis when it comes to how much we should be taking, how it impacts the overall plan, but let’s see where it takes us. Okay. You are choosing to spend more from investments early in exchange for higher guaranteed income later, lower longevity risk, better survivor protection. It’s the very favorable trade.

Okay. So I, so this is a very, very important section right here. So what it’s doing is it’s raising the trade-off. So this is one of the most important things that a retirement planner can do is make the trade-offs of the decision that you’re currently considering make those trade-offs visible. We have to raise them to the surface.

What I don’t like here is that it’s telling you that it’s a very favorable trade without any type of facts or analysis to really back it up. It’s just simply asking you to trade. Ultimately, trade-offs have positives and negatives. So it’s only going through the positive side. So it is raising it in the sense that you are going to spend more from your investments early.

So technically, I guess that is kind of the trade off in exchange for more income later, lower longevity risk and better survivor protection. And it’s telling you it’s a very favorable trade off. But again, I need to see some numbers. I need to see what are the consequences. What is the sensitivity analysis? So I love that it’s bringing up this trade off, but it’s asking me to trust it. And that’s, I need a lot more than that.

Okay, so now we have the exact action plan, 62 and 64, everything that’s basically covered. This has changed now 80,000 to 95,000 from the portfolio, 50 to 100K from the IRA. This is also adjusted from up above. I believe it was not much, 40 to 100K. Asking you to monitor your brackets carefully. Okay, it’s giving us all of the reasons why here.

What the AI missed: Taxes, healthcare, and asset basis

Long story short here is that it is telling us that we should take Social Security now for the wife, delay the husband’s Social Security, use that window while income is low to do Roth conversions, pull from the brokerage account first, then pull from the IRA to supplement any additional income needs, and then do Roth conversions from 40 to $100,000. Okay, so.

If this video was simply about this planning case, I would really like to dive in and ask it lot more questions, but instead I just want to raise a couple of things that I noticed. So first on the spending, it’s not taking into consideration the taxes on withdrawals. It’s telling us to pull out somewhere between 80 to $100,000. Well, the spending goal is a little more than $100,000 a year.

There’s only so much in non-qualified cash that we can pull from without paying taxes on. It mentioned that there may be, or capital gains need to be managed. There’s 250,000 inside of the brokerage account here. So do you have a lot of high basis stocks where there’s very little tax friction, where you can sell them and have no tax consequences? Or are they low basis stocks that you’ve had for many, many years, and if you sell them, you’re going to create tax consequences, which could alter the plan.

This is something that we would absolutely ask and need to know before we could make any type of recommended action. And it seemed to have just kind of glossed over that and made the recommendation on what you should do. Additionally, it doesn’t ask anything about healthcare. Healthcare is a big part of retirement. Those expenses increase for most people as you age and the spending itself. Is that inclusive of healthcare? Is that not inclusive of healthcare?

For me, if somebody came in and they told me, Troy, I want to spend $8,500 a month, that should be on average what I’m spending. Typically, that’s what they want to spend to live their life. It doesn’t include healthcare. Now, before I show you what our software came up with, I want to be fair to AI here. This is a very thoughtful answer.

It covered withdrawal sequencing. It looked at taxes. It looked at Roth conversions. It looked at a lot of items that I really didn’t even ask it. And it gave an answer that for many people, they would say, you know what? This is a really good answer. I’m pretty good here. I’m going to stop.

Professional software analysis: The 61% probability of success

But that is what makes this next part so important. If you’ve followed this channel for some time, you know we do a lot of case studies where we look at this particular scenario and many similar to it. And one of the things that we always get to is the probability of outcome. What’s the likelihood of success?

So here we go, we click this, we run 1000 different simulations of this exact scenario. So we’re not even looking at any of the Roth conversions, we’re just looking at all of the parameters at the base case. So the spending, the ages, the longevity, the assets, the social security strategy.

And immediately off the top, in 610 out of 1000 scenarios, this couple passes away with money. But you have to ask yourself, is that good enough for you? 61 % probability of success, is that good enough? Now, there are some limitations with this Monte Carlo analysis and any Monte Carlo analysis is that it’s just a snapshot in time.

This is going to change over time based on the decisions that you make in subsequent years, how the markets perform, and of course, what you’re doing from a Roth conversion standpoint, many different factors. So one thing I do want to point out here though is look at the first 10 years of retirement roughly to 2035 here.

So even in the scenarios where we run out of money, about 10 years from now, we still have a very tightly wound group of numbers here. You may still have around a million bucks at this time 10 years later and feel pretty good, but your trajectory, the path that you’re on is leading you to run out of money. But for most of us, we probably still feel pretty good at that point.

So the takeaway there is that mistakes in retirement they compound over time and you aren’t able to initially see them typically within the first five to 10 years of how they’re going to play out over a long period of time because the difference between a good outcome and a bad outcome oftentimes is very, very narrow in the first 10 years or so.

It’s after that year 12, year 15, year 17 that you really see that divergence of outcomes based on decisions that you made 12, 15 years ago. So this is what we talk about when we say you have to compound good decisions year after year after year after year because it’s not like making a bad investment, right?

It’s not like any accumulation phase. I go out and I make an investment in a timeshare. Okay, I know pretty soon that that’s probably not a good investment. I buy Apple stock, let’s say, it crashes. Well, that tells me pretty quickly that that was a bad decision or I lost money.

In retirement, decisions don’t show up for many, many, many years down the road, the outcomes of decisions. So just keep that in mind.

Social Security deep dive: Comparing 6 different strategies

Okay, now here’s the real interesting part here because it gave us a very good answer about social security. So this is a social security analysis where we’re looking at the answer that ChatGPT gave us.

So she taking it at 62, he at age 70, and comparing it to multiple different options. So both of them taking it as soon as possible at retirement, at full retirement age at age 70, he at 70, and she at full retirement age, and this actually gives the overall plan the highest probability of success outside of number six over here, which is he takes it at 70 and she defers until age 67.

And we look at the total lifetime benefit, $1.5 million from Social Security by going with the strategy, the Social Security strategy that CHAP GPT recommended compared to scenarios two, three, four, and five where they take it as soon as possible. So now, 1.275 million, 2.74 in lifetime income. Same at retirement. At full retirement age, 1.436 million. And if they both delay until age 70, only $5,000 more per year.

So this scenario six gives us the highest amount of lifetime income from social security if they make it to life expectancy and the highest probability, but it’s not much more. And so in this scenario, looking strictly at the social security analysis, this is probably what we would recommend for a couple of reasons.

One, the math is clear enough to make a lot of sense, but also it comes down to a situation of what Chachi PT pointed out is the longevity insurance. We really wanna make sure that higher earning spouse has delayed as long as possible and there isn’t much benefit to the younger spouse, or excuse me, the lower earning spouse to delay because if something were to happen to one of them, that Social Security check goes away entirely and they’re only left with a higher earning spouse.

Why a “good decision” isn’t always the “right answer”

The answer actually isn’t bad. So even though the Social Security answer is very solid, it’s not the right answer in the context of what’s the right thing to do for the overall plan. So before we get into some of the key planning considerations here and even the Roth conversion window that they talked about, it’s important to recognize that you can make a very smart strategy selection, so make a good social security decision inside a plan that still doesn’t work.

So it’s a good decision, but it’s not the right question. It’s not the one that really matters. So now I wanna look at the tax conversion strategy. So when we enter the case into one of our tax planning softwares, we notice a couple of things. One, very similar, very low estimated ending balance, and this is just one of the Monte Carlo simulations.

But compared to the base strategy, which would be no Roth conversions, we do end up with a little bit more money than over here. This projects that we end up with zero. So the Roth conversion itself can make sense, but the bigger question is, are you comfortable later in life seeing the accounts dwindle to this extent?

Here’s what the actual Roth conversion schedule would look like from an advanced planning standpoint, meaning looking at all of the context of the entire case. And we can get about four years of smaller conversions in here. So a couple of things. So GPT was right in the sense of giving you a range of $40,000 $100,000.

And it looks like for about four years. So at that point, the non-qualified is most likely run dry. And the benefit of converting more dollars to Roth in that window pretty much exhausts after 2029. So continuing to convert beyond that would have a negative expected outcome when we look at total value over the course of your retirement.

Okay, so we have a clear discrepancy between what both of the retirement planning softwares that we use for our clients and what Chatsy BT is showing us. And the biggest discrepancy is the likelihood of success over long periods of time. But why is that?

AI’s portfolio projections at age 95

So what I want to do now is look at two different things. One, I want to ask Chatsy BT the same thing that we asked our probability modeling softwares, how much money will be left at the end? What is the probability of success?

Then I want to ask Chatchie BT, what did it overlook? Because as I noted earlier, it’s not a matter of, because as I noted earlier, there were many things that were left out of their recommendations, such as taxes on the withdrawals, healthcare expenses, and many other items that simply were either glossed over, simplified, or just completely left out.

So I have provide portfolio value based on the entirety of your recommendation at age 95 for spouse. So I asked it to provide the portfolio value based on the entirety of your recommendation at age 95 for the spouse. So the wife is still alive at age 95.

So first it gives us some assumptions that it’s used, starting portfolio value, spending need 102K after tax. Again, but the entire analysis never gave us 102,000 after tax. I think it was 80 to 90,000 from the portfolio. It has the social security strategy. It assumes an average annual return here of 5%, 2.5 % COLA, longevity expectations, higher withdrawals.

So this is what I was just talking about, 85 to 95,000 a year. They simply would need to be hired to meet that after-tax spending goal, but that’s okay. plus to pay taxes on the Roth conversions. So portfolio at 70, 1.2 to 1.3. This was fairly in line with what we had in some of our scenarios. could have been somewhere between, a realistic outcome would be 700,000 to 1.5.

Age 70 to 90, so at age 90, 1.3 to 1.6. And then wife age 95, about 1.1 to 1.5 million remaining. So you would walk away from this feeling pretty confident that the social security strategy is good, that you should do Roth conversions all the way until age 70, and that you would not run out of money.

Okay, it gives us some insights here, things to consider. The bottom line is very high probability of never running out of money, strong survivor protection throughout, and you end retirement with between $1.1 and $1.5 million.

10 critical factors AI overlooked (IRMAA, inflation, and more)

Okay, so now what I did is I said, what did you overlook in this exercise? And I think this is a very key question because as I noted earlier, there were many things that we identified that were overlooked.

First thing, taxes were modeled too lightly. So this is the biggest issue. It did not take into consideration taxes at all. Federal tax brackets, social security taxation, capital gains stacking with ordinary income. When you sell stocks that have gains, it fills up the lower part of your brackets and pushes Social Security and other sources of income like Roth conversions into higher income tax brackets.

Any potential IRMA surcharges on Medicare future require distributions. This could impact Roth conversion sizing. You might under convert, leaving larger RMDs later. So basically the whole picture. So the difference between good and optimized tax planning here can easily be $200,000 to $500,000 over your lifetime. That’s very accurate. Depending how much money you have, often we see $500,000 to a million plus in potential tax savings over a lifetime.

So that’s pretty accurate for this net worth range of about $1.4 million in assets or so. Number two, sequence of returns risk was smoothed over. They just looked at a steady 5 % return. And that’s a big one. That’s a big one. Spending was assumed too smooth. In the goals, we said higher spending in the first 10 years, but they didn’t fully model any of these relative to the primary objectives.

Number four, healthcare and long-term care costs two different components. So healthcare is obviously out of pocket, not included by Medicare. So your premiums, your Part B, your Part D, any out-of-pocket expenses, copays, et cetera. And then long-term care, completely separate. Medicare doesn’t cover long-term care except for 90 days of skilled nursing facility and you’re on your own.

Number five, surviving filing status shock? So this is a big one. I’ve never heard it called this before. We call it tax bracket compression. So when one spouse passes away, you go from the married filing jointly to single brackets, almost everything cuts in half. Just did a video on this. So that’s actually very, very good timing.

Highly recommend that you check that video out. We’ll put a card at the end of this video so you can see that one. But almost everything does basically cut in half. So if one spouse passes away prematurely, not at 90 and 95, an entire plan can be up in.

Inflation was simplified, didn’t differentiate between healthcare inflation and lifestyle inflation early on. They didn’t know Monte Carlo or probability modeling, so it was a single path estimate. And it gave you a high degree of confidence with a single path estimate. And it’s saying here there could be, even with a $1.3 million ending value, there could be a 20 % chance that you have less than $500,000 and a 20 % chance you have more than $2 million.

So that’s exactly what we showed you. Behavioral risk. this is the, I say it all the time, the number one value a good retirement planner provides is simply being there to raise trade-offs, to have discussions, to make everything visible to you so you can make good decisions. But then it’s managing the behavioral risk over time. This is a big one. It’s just simply a big one.

The one thing that no math and no retirement plan and no amount of modeling can ever quantify is the risk that as we get older and things change as far as our willingness to take risk, our capacity to take risk, how do we behave under circumstances of extreme pressure and stress? How do we make decisions when we’re older and these things are as consequential as possibly running out of money, not having enough for healthcare? This is a big one and it becomes bigger and bigger the longer you live.

Social security policy risk, asset location and withdrawal efficiency. This is a big one, it’s something I pointed out earlier. Which accounts to draw from, tax efficient sequencing. So, then it goes to the most important takeaways here, so I think we’ve kind of covered it.

Final takeaways: Managing retirement as a complex system

The big takeaway here is not that AI is bad at social security, it was actually very good at social security, but it’s that retirement itself is a complex system. And when you’re making decisions, understanding that every decision that you make interacts with every other aspect of your retirement far more than it did in the accumulation phase and that your retirement needs to be managed as an entire system, not individual parts looked at in isolation because you can have a really good answer to an isolated part, but either you’re asking the wrong questions or you don’t have the full picture.

hopefully by going through this and seeing how I look at this, seeing how I would bring this up, from a retirement planning conversation has helped you to either ask better questions, to understand the context within the greater picture, and hopefully help you have a better retirement. And for the bracket compression video, when you lose one spouse, some of the biggest risks that you encounter, click on this video right up here. I recently did it. It’s a great video and it’s gonna help you understand that in a lot more detail.