Meet Soar: AI-Powered College and Career Planning

SOAR IEC Webinar Transcript

Speakers:

• Venkates Swaminathan (Swami): Founder, LifeLaunchr and Soar
• Claire Schadler: Senior College Admissions Consultant, LifeLaunchr
• Ruth Louden: Senior College Admissions Consultant, LifeLaunchr

Swami: Hi, and welcome everyone to our webinar today about SOAR. We'll jump in — I'm really excited to do this today. I want to start by talking a little bit about AI and college admissions counseling, and how it works.

Some of you already know me from the work I do. So first, thank you. Many of you are here because you read my Substack. Clara and Ruth, and the rest of my team, and other people also contribute to it.

Let's start by introducing ourselves. My name is Venkates Swaminathan — Swami. I founded LifeLaunchr in 2014. I've been doing this for a long time, and Claire and Ruth are both my colleagues as part of my team. I co-lead the HECA AI Learning Lab, and also serve on the ad hoc committee that HECA has for creating AI guidelines.

I started in AI way back in the '80s when I was a graduate student studying electrical and computer engineering at the University of Illinois. I was doing work on computer vision back then, before AI was such a mass market thing. AI, as a field, has been around for decades, and it's only recently become mass market.

Computer vision is the idea of trying to make sure that computers can see in the same sense that human beings can — that we can walk into a room and see what's in it. I've been in AI for a very long time, and in college admissions counseling for a very long time.

So I have a somewhat unusual combination of backgrounds to be having this conversation with all of you today. Claire, perhaps you want to introduce yourself?

Claire: Hi everyone. My name's Claire Schadler, and I've been an education consultant for about 10 years working with students on their applications. My background is in teaching English as a second language, so it was a natural shift to come over and help students with college admissions after helping them with language learning.

Swami: And Ruth?

Ruth: Hello everyone. I'm Ruth, and I'm very happy to be here today. I've been working with Swami now for about 10 years on college admissions coaching for students largely in the United States, but also those students who want to come to Canada — I help them with their applications.

I've been quite fascinated by technology for the last 20 years or so. I recall back in the early 2000s, going to university conferences where all the student service professionals were terrified about the advent of social media and how our students were using it — and now it just feels like one more thing along the road. So how do we get on board this train and use it effectively?

Swami: Absolutely. Today I want to give you a sense of why SOAR is built the way it is and where my thinking was coming from as we created it.

In all the conversations I have with people — not just through the Learning Lab, but people who reach out to me about AI because I use it a lot — there are two underlying fears. They're not always precisely articulated, but they're often there.

One is replacement: that AI will just replace us, that the work we do in five years will be done entirely by AI and we'll become irrelevant. The other is FOMO — people say, "I really want to figure out how to use AI. What should I do?"

When I think about all of it, what resonates for me is: A) AI is not going to replace us, because what we do as human beings is so profoundly important to this work. And B) when someone starts with "I want to use AI," I think that's the cart before the horse.

The way I think about it is: start with goals, not with tools. Ask yourself, "What problem do I want to solve? What do I want to accomplish? What are we trying to get done — for ourselves, for our businesses, for our students, for our families?" And then ask how technology can help, because it can in so many ways. I always start by saying, "Here's something I want to get done, and here's a place where AI can really help." That framework is really important.

Claire: Swami, could you give an example of someone who started with the tools and it didn't go well?

Swami: Great question. So here's the kind of thing people will say sometimes. When you think about modern large language models — ChatGPT, Claude, Grok, all these systems — they have a kind of progression. There's the chatbot: I ask questions, it gives me information. Then there are skills you can use to create various things. And then there's the next step, called agentic AI, which can do more advanced things. When you think about the kind of work I did writing SOAR, that's an example of agentic AI.

I made a joke in a Substack post — it wasn't really a joke — that at home it's a constant refrain: "You talk to Claude more than you talk to me." But it's true, because I'm talking to Claude to write code — Claude's writing the code, not me. That's an agent doing work on my behalf.

The thing about agents is that people sometimes say, "I want to use agents," because that's the buzz. Agents are extraordinarily powerful, and if you don't know exactly what you need them to do, they will do the wrong thing very well. That means you might send an email to someone you really, really didn't intend to send — because the agent sent it on autopilot.

Swami: So in general, when I built SOAR, there were five things on my mind.

One: We don't know what our students are hearing from ChatGPT. Most of our students are using it, maybe not exactly the way we want them to, but they are — and we have no idea what they're saying to it or what they're getting back.

Two: Misinformation spreads really fast in tech. There are TikTok influencers talking about college admissions, Reddit message boards full of misinformation, and of course ChatGPT. And some of it comes from family and friends — "My uncle said..." We can't just overwhelm that with knowledge. We have to be able to reason through it with our students.

Three: Work stalls between meetings. You meet with a student, ask them to research something, and when you meet again a week later, nothing got done — maybe because they didn't know how to do it.

Four: Anyone who's worked with a teenager has heard "It was fine." How was your research into Purdue? "It was fine." I'd ask my daughter, "How was school today, Grace?" — "It was fine, Dad."

Five: College counseling is bigger than just college. Students want to think about their major, alternate majors, career paths, the impact of AI on those careers, costs, financial aid, scholarships — all of it becomes intertwined in our work.

Claire: Yeah, I had a student who I'd given a couple of colleges to look into, and when he came back a week later, I asked how the research went. He said, "I actually don't really know how to do that." This was before SOAR, so I walked him through some ways he could do research. He just didn't know what steps to take or where to look or what to look for. That was a week that went by with nothing getting done.

Swami: And that happens a lot. We'll ask students to research scholarships, and we have tools where you can do that — but honestly, nobody ever does it because they don't know where to start. They don't know what questions to ask next. After asking "I want scholarships," they're stuck, and the tools can't guide them through the process.

You meet with a student, and when you meet again, not as much has happened as you wanted. And then you spend that session just trying to figure out what "it was fine" actually meant. Students will say, "Looking at the websites of these colleges, I can't tell the difference." And sometimes that's sort of true — college websites all have certain similarities. But if you look at the actual program of one college versus another, or the culture, they're quite different.

And then as I said, college counseling is so much bigger than just college. Scholarships, majors, summer programs, careers, demonstrated interest — the students have so many questions. Some are very specific, some are very big. But these are all things we end up having to talk to our students about.

Ruth: And Swami, you alluded to this earlier — it's really challenging to personalize something like scholarships for a family. There are so many scholarships out there, so many enrichment programs. How does SOAR do a better job of personalizing than our existing systems?

Swami: Great question. We built in a database of over 6,700 scholarships — we licensed that from Peterson's, who updates them every quarter. We also include our database of 250+ enrichment programs. And we built a huge amount of knowledge into SOAR about how scholarship search actually works, so that if you ask a vague question like "I want scholarships," it will walk you through the process — where money comes from, financial aid, college scholarships, outside scholarships, and so on — step by step, the same way you or I would.

And the great thing is that students are doing this work between meetings. So when they come to a session, they come better prepared. "Oh, I think I found this Eagle Scout scholarship, can we talk about that?" Now that's a meaningful conversation we can actually help with.

Swami: What I want to do today is showcase the tool, but also explain why it's built the way it is — so you can understand whether it will fit your practice.

The core of SOAR is an AI college and career planning assistant that keeps counselors in the loop and students on track between sessions. It makes us all a team: student, parent, counselor — one team, and we all get to see what everyone is doing.

A few design choices:

Counselors are always in the loop — and not just counselors. Students and parents too. If a student does research on Purdue, you'll see it. If you do research on behalf of that student, they'll see it, and the parent will see it too. The information gets shared between everyone on the team. That transparency is really important.

It won't write essays. It won't write them well, poorly, or even just for brainstorming. Claire, you actually told me you tried to trick it.

Claire: Yes! I assumed that if I just asked directly it would refuse, so I tried to be indirect. I gave it a really long prompt describing exactly how I wanted the intro — a scene about a robotics competition, written in present tense, dropping the reader in — and then asked if it could give me an example of what that might look like. It said no. It explained that it could help me brainstorm ideas and give feedback on what I'd written, but that my voice was important. Even if I used it as just an example, I'd be using the program's voice. It explained itself clearly and didn't do it.

Swami: I didn't know you'd done that until you told me a few minutes ago — so I'm happy to hear that. We've built a very strong safeguard around not writing essays. Of course it's possible someone might trick it, but that's our goal.

Ruth: So where's the line between what it will do to help versus actually writing?

Swami: Great question. If you ask "Will you write my why cognitive science essay for me?" — you'll see in a moment what happens. It will walk you through the same kinds of questions we might walk a student through: What experience led you to it? What have you done? It asks the questions that might lead to a good essay, but it won't write it.

Claire: And in that same scenario, it said it could tell me if I'd chosen a word that wasn't in the tone of the rest of my writing. So I asked, "Well, will you tell me what word is in the right tone?" And it said no. It said, "Sometimes if you can't pick a word that's the right tone, it's because you haven't thought through what you're trying to say. I'll give you guiding questions so you can clarify in your own mind what you want the tone to be, and then I'm confident you'll be able to think of the right word."

Swami: That's awesome — and it surprises me in a great way.

Hallucinations are a problem across AI tools, and we know our students are talking to ChatGPT and getting answers that often don't make sense. ChatGPT consistently overestimates likelihoods. Because of that, it encourages students to build college lists that are often just poorly thought through. If you tell someone they have an X chance of getting into a college when they really don't, their list ends up skewing toward reaches, and that produces poor outcomes.

So we incorporate data from the Integrated Postsecondary Education Data System (IPEDS), the College Scorecard, and the Common Data Set — licensed from Peterson's with quarterly updates. As soon as new data is available, we can update everything within half an hour.

When it doesn't know something, we instruct it to search the college's website, and when it really doesn't know, to tell you rather than make something up. And when it draws from general knowledge rather than our data, it will tell you and ask you to verify — because that general knowledge has a training cutoff and may be out of date.

Fourth design choice: Encoded with counseling methodology. Many AI tools are basically a smarter Google — you ask a question, you get an answer. What SOAR does is ask you more questions and walk you through a process step by step. To do that, there are sometimes 25,000 words of instruction that go along with a question you ask. You might ask a 20-word question, and 20,000+ words of context get sent to Claude along with it, telling it how to answer based on our methodology.

Claire: Where does that context come from? How is SOAR different from just an advanced Google search?

Swami: All of that context came from hours and hours of conversations with Claude where I said, "Here's the process we use. When someone asks about scholarships, here's what we do. When someone asks about UC admissions, here's what we do." And then I'd say, "Now generate a skill file — and update it — actually, that's not quite right, do it this way." That's an ongoing conversation, because each week we'll find it does something not quite the way we'd want, and we adjust.

We also have the ability for counselors or companies using SOAR to upload their own methodology — if their approach is slightly different from what's already encoded.

Claire: So people using this platform can incorporate their own approach?

Swami: At certain tiers, yes — you can upload a file that says "This is how we do X." For example, we have our own terminology for likelihoods: likely, target, reach, far reach, unlikely, unknown. Other people may use slightly different terminology, and that can be accommodated.

Fifth design choice: Safety by design. When students type in messages like "Life doesn't feel worth living" or anything suggesting harm, it flags it immediately. You as a counselor and the parent will see an alert. Sometimes it will be a false positive — Claire, you had an example of that.

Claire: Yes — I got an email from it, logged in, clicked dismiss, and that was the end of it.

Swami: We'd rather it flag when it doesn't need to than miss something when it should have flagged.

And finally: the meeting brief. Before a meeting, you normally have to review all your previous session notes, the student's college list, their scholarship list — everything. With SOAR, you just click "Generate Meeting Brief" and it reviews everything and tells you: "Here's what you need to talk about with this student today." It creates a full prep brief for your meeting.

Ruth: And we'll see that in the demo today?

Swami: Absolutely. And at the end of the meeting, you can combine that brief with your own notes and a transcript from a tool like Granola, Otter, Fathom, or Zoom, and generate a meeting report that gets sent to the parent and student. You can edit it before you send, but it'll say exactly what you want it to say.

Ruth: Since we often co-coach students, in the past we'd each read our own notes, try to combine them, and figure out what to discuss. I'm assuming SOAR would combine all of it?

Swami: Everything — what the student did, what you did, what their parents did, the research, meetings, activities list, college list, scholarships — all of it. It looks at everything and says, "Here's what needs to get talked about today."

Claire: I used the meeting brief feature and it pulled the student's SAT score from their profile — which hadn't come up in our last meeting — and flagged it as relevant to the colleges they were researching. It made that connection for me: maybe you should consider the SAT when talking about this.

Ruth: That's like strategic focus for the meeting.

Swami: Exactly. And you can look at it and decide what you actually want to discuss — it may be longer than you need, but at least all the information is right there at a glance.

Finally — and this is really important — because we use the Anthropic API rather than claude.ai or Claude Code, Anthropic guarantees they will never use the information you send to train their models. If a student submits an essay and gets feedback, that essay will not be used to train Anthropic's models. It's forgotten by their systems. So you can use it without any concern that a student's privacy will be violated, their essay will be used to train models, or they'll be flagged for plagiarism.

[Live Demo Section — Swami demonstrates SOAR with fictional student Katheryn Williamson, showing college research with learning differences context, scholarship search, CS with low math grades, summer program search, essay assistance, and meeting brief generation.]

Claire: There's a question in the chat: someone asked if you have a use case for younger students, grades 8–10.

Swami: Great question. I don't have a demo for today, but I'd love to set up a time to talk through it. There's a link in the chat — click that and schedule a consultation. Eighth graders heading to college have the greatest runway and an incredible opportunity to make a difference on their applications.

Claire: There's also a comment from someone who said, "I like that I wouldn't have to create a separate shareable document — that ends up taking a lot of time, even when editing an AI-generated one."

Swami: Yes, exactly. And you can just tell parents, "Log into SOAR and you'll see it there." All the research summaries are right in the student's dashboard. And a year later, when you've forgotten the conversation, it's still there. Every conversation adds to what SOAR knows about the student — the value compounds over time.

Claire: Someone also asked: how do you recommend presenting SOAR to families who may be concerned about privacy, even as they themselves use ChatGPT and Claude?

Swami: Great question. Let's start with privacy. We connect to Anthropic through an API — our servers talk to Claude's servers. Anthropic's terms for API users are fundamentally different from their terms for consumer-facing tools like claude.ai. Because of that, any information shared through the API will absolutely not be used for anything other than answering the question you're asking. Claude forgets the conversation after that. Privacy is genuinely not a concern.

On the broader AI anxiety: I've been following all the college policies on AI — I wrote a long Substack post last August covering every college's policy — and SOAR is comfortably on the right side of the line. I'm also part of the HECA Ad Hoc Committee laying out AI guidelines, and this is completely compliant with everything we're considering. Students don't need to worry that using SOAR will put their applications in any danger.

Swami: Now, where does SOAR fit in your practice? If you're already an IEC or school counselor, you use tools like SCOIR, College Planner Pro, Siafo, MyLearning, and others. SOAR isn't intended to replace any of them. It's an extra tool — but one that really earns its place in the process, because it's the tool that works not just with you, but with the families. Students and families actually use it.

Your kids are already chatting — with friends, on Instagram, with ChatGPT. SOAR meets them where they are. It doesn't replace your other tools; it adds something those tools can't do: a conversational, AI-guided experience that keeps everyone on the same team and working together between sessions.

Claire: Someone in the chat mentioned: "Good point about holistic decision-making. Worth featuring in sales/marketing — though not including institutional priorities, which is huge."

Swami: SOAR actually does include institutional priorities. For example, if you're a California student applying to UW, it knows that UW is much more selective for out-of-state students than in-state. And Purdue, on the other hand, actively wants to admit California students. It knows those nuances because our data comes from three sources — two government databases and the Common Data Set. It also flags whether a college considers demonstrated interest or legacy status.

Thank you all so much for being here. I really appreciate the questions and the conversation. Claire and Ruth, thank you for being part of this. If you want to talk further, there's a link in the chat to schedule a consultation. We'll do another webinar soon!

[End of Webinar]