Long Story Short
AI isn’t the easy way out; nor is it the answer to every problem. To truly leverage it, start by identifying a real friction point you want to solve, what it’s costing you, and what steps in your process should remain human. THEN go see what’s out there. We pulled Grace Kim‘s method into the Tech Adoption Toolkit if you want to start solving the friction you’re living with today.
For years, I had strong opinions about technology decisions. Build or buy? How much should we control the way a team uses a tool versus letting everyone go wild? Where is it worth the slog of change management, and where do you just need to get the thing done? I knew how I liked to answer those (buy!!!) — and often answered before someone finished asking.
Someone asked me “build or buy?” a few months ago. And I legitimately had no idea!
It’s not that I forgot my predisposition toward buying. It’s that AI has shifted so much, so fast, that I don’t have confident answers to any of those questions anymore — and I’m not sure I should. (If you do, I’d genuinely love to hear how.) It’s a new world, and it is SO easy to get lost in the pressure to be doing something, the volume of what’s suddenly possible, the worry about privacy and ethics and whether you’re getting it wrong and ruining the earth in the process.
But here’s what’s helping me stay steady right now: when I get lost, I go back to what made me sure in the first place. Underneath all those old opinions was always the same thing — getting things done matters, and grounding in where you actually are matters most.
Here’s what that looks like for me right now:
- One org we work with has Salesforce as their long-term vision. AND they needed a working pipeline now — so we built an Airtable in about two hours with AI and they ran with it happily for eight months. Not the perfect system. The one that got the work done.
- AI could absolutely do my invoicing. AND I like sitting with the hours for Helia’s partners each month — it’s an honest check-in on where we are, and it’s worth the hour to me.
- Transcripts are great and I use them. AND I still take verbatim notes the entire time in every Helia interview, because it helps me stay present in the conversation and the information lands differently in my brain. (Meanwhile I take shockingly few notes in brainstorms.)
None of those are about the tool. They’re about knowing what the work actually needs, and where I actually am.
That’s the part that matters. AI isn’t the easy way out — it’s another powerful foundation, with more tools hanging off it than any of us can keep up with. The folks who get it right aren’t the ones who pick the perfect one. They’re the ones who start from where they actually are — their real friction, their real data — AND build toward where they want to go, in lockstep with the people doing the work.
The person who builds from the inside out
Grace Kim came to Helia through Collective Member Mike Chae. She’s been doing this work longer than most of us knew there was a this to do. She started as an earth scientist, with a PhD and years at NASA working with satellite data, asking big foundational questions about how a system actually works and what the evidence really shows. From there she moved into data science at a large consulting firm, and then went out on her own.
We could tell who Grace was when she talked about why she made that move. In big consulting, the minute new technology shows up, you get a five-person team — a strategist, an engineer, a project manager — and a price tag that climbs fast. Which means the organizations who get real help are the ones who can already afford it. Grace went independent to bring that same level of thinking to the rest of us. (Not everyone can afford a 3–5 person consulting team — and maybe we don’t all need one!) That’s a very Helia approach to the world and we love it!
What she said, over and over, is that the right technology decision starts from the inside out — not the outside in. Not “here’s the shiny thing everyone says you need,” but “here’s the friction you already live with.” The annoying task. The question you can never quite get a straight answer to. The thing that’s just slow. Grace starts there — with what’s already not working for you — instead of with whatever’s being sold.
My favorite story of hers is the medical billing one. A practice was losing money on claims slipping through the cracks and couldn’t even see where, because their software couldn’t answer the question. So Grace built something genuinely sophisticated: she pulled the data their system wouldn’t easily give up, organized it around how the team actually worked, mapped where each payment was supposed to come from, and used an open-source AI model that kept all the sensitive information on her own machine instead of sending it off somewhere.
And then the thing she handed back was an Excel sheet. Because that’s how the team already worked. No one had to learn a new system or do more work to use the very thing built to help them. In Grace’s words, “if a new tool doesn’t talk to your existing system, or it makes you do more work to use it, then it’s just solving one problem by creating another.”
That’s when I knew Grace would get things done. The build was advanced and the landing was right where the team already was — start from the real friction, design around how people actually work, and don’t make anyone leave where they are to get the benefit. It’s exactly what Helia folks tend to want: get in, start right where you are, and help the work get to the next place.
The Key Takeaway
AI isn’t the easy way out; nor is it the answer to every problem. To truly leverage it, don’t worry about picking the perfect tool (it’ll change by next quarter anyway). Start by identifying a real friction point you want to solve, calculate what it’s costing you, and decide which steps in your process that should remain human, THEN evaluate whether to build or buy.
If a new tool doesn’t talk to your existing system, or it makes you do more work to use it, then it’s just solving one problem by creating another.
Grace’s process
If you’re a “just show me the template, I’m ready to go” kind of person (hi, COO Libby), we pulled Grace’s method into the Grace’s Tech Adoption Toolkit so you can map your own process and start finding the friction today.
If you’re a “tell me why this works first” kind of person (hi, CEO Jess), read on and notice what’s not in it: a single recommendation about which tool to buy. That’s on purpose. Grace’s method is the same every time, no matter what’s new this month.
- Step 1: Map your process. Pen and paper are fine. In fact, it’s even better if you can close your computer and take some time away from all the notifications for a little while to work on this. Walk one process start to finish — intake, service delivery, your financial workflow — and get specific about every step.
- Step 2: Layer on your tech. Go back through and note what actually runs each step today — the form, the spreadsheet, the email, the phone call.
- Step 3: Mark what should stay human. Some steps are better slow, private, or personal — on purpose. Name those before you automate anything.
- Step 4: Find the friction and price it. Where is a step slow, breaking, or quietly costing you? Put a number on it — hours or dollars. That number is your benchmark for what’s worth spending.
- Step 5: Then decide — build, buy, or wait. Now the hard questions (build vs. buy, early vs. late, how much to take on) have something to stand on: your real map, your real costs, and where you sit on the adoption spectrum.
The order here matters — a lot. By the time you get to the tool, you already know exactly what you need it to do.
How Grace’s Process Works
Step 1: Map your process — pen and paper, start to finish
Grace’s first move is low-tech. Sit down and talk through one process from start to finish: “Can you get on a call and talk through how you deal with intake for a client, or how a client moves through service delivery from start to end?” It can literally be a drawing — “just pen and paper, what are the key steps for what happens, start to finish.”
As you go, she wants you to notice one thing: “think about the steps that are really benefiting from human involvement, and which ones can use help from a computer.”
The honest hard part is that this asks you to think in workflow mode — a real cognitive switch to think more algorithmically. It’s the piece people find hardest, and it’s worth not rushing.
→ If mapping your own process feels like staring at a wall, this is exactly the kind of thing Grace does with teams. Book 30 minutes with Grace.
Step 2: Layer on your tech — what actually runs each step today
Once the steps are down, you go back through and map the technical tools, software, and the apps you’re using to make all of that process layer happen.
For the doula collaborative Grace works with, that looked like:
- the intake form is a Google form
- the template email goes out through Gmail
- the progress tracker is a Google sheet,
- the intake interview is a phone call where private information gets shared (and never transcribed)
- determining eligibility and referrals is — in her words — “no tech, pure human genius, and it’s best left that way”
Step 3: Name what should stay human — on purpose
Here’s where Grace pushes back on the pressure to automate everything. Some steps are better slow, private, or personal, and you should decide what those are before you reach for a tool.
Her clearest example is money. “There are certain tasks that benefit from being slower. Approving invoices is a great one to be slow, intentionally, because you want to make sure all the money going out is legitimate.” (She’s seen scam invoices make it all the way to final approval.) For anything financial or high-accuracy, she asks: “where do humans really shine, and where does accuracy matter?”, and then builds in a verification layer, or keeps it fully human.
The doula intake interview is the other one. It could be a form. But it’s “a huge opportunity to educate the client about what a doula is,” AND it’s where someone shares sensitive things like health insurance eligibility — “information we don’t want on a form, that we want to keep very protected and private.” So, they don’t automate it. “That human connection is such a valuable part of the process.”
This is also where the privacy and ethics worry quietly gets answered. You’re not deciding it in the abstract — you’re deciding step by step, right where you can actually see the stakes.
Plenty of steps are great to hand off, of course. For the doula team, every spot where someone was copying and pasting — drafting the same intake email, retyping a name into a tracker — became an automation. “Anytime there’s copy and pasting” is Grace’s tell that a step is ready for help.
And it’s worth zooming out once in a while: if a task disappears into AI entirely, does anyone on your team still learn to do it? Sometimes you keep a thing human not because a tool can’t do it, but because the skill is worth keeping in the building.
There are certain tasks that benefit from being slower. Approving invoices is a great one to be slow, intentionally, because you want to make sure all the money going out is legitimate.
Step 4: Find the friction — and put a number on it
Now you go looking for where it hurts: where is a step slow, breaking, or quietly costing you? Grace’s medical billing project started exactly here — the practice was losing money on claims slipping through the cracks and “literally couldn’t answer that question with the capabilities of their existing software.”
Her move is to make the cost concrete. “Can you quantify, in dollars or hours, what it’s costing you? Now you have a benchmark for how much you might consider spending on fixing it.” Once you know a problem is real AND you know what it’s costing, “you actually have a solid number to work from on what’s worth it for you to build or buy.”
→ When the friction is real but the fix is genuinely complex — like pulling data your software won’t give up — that’s custom-build territory, and it’s what Grace does. Book 30 minutes with Grace.
Step 5: Then decide — build, buy, or wait
Only now do the hard questions have something to stand on.
On build vs. buy: the cost of building has dropped and a tightly scoped vision might even mean a freelancer can do it. AND Grace still asks a few timeless questions:
- Who’ll actually run and maintain a build?
- Will the vendor still be around in five years?
- Does it make the people who use it every day do more work, or less?
And sometimes the answer is neither. The best option can be a capability that shows up inside a tool you already use. When Grace was helping the doula team, the automation they needed appeared right inside the Google workspace they already worked in every day — nothing new to adopt, nothing new to learn, nothing sitting off to the side. She could just teach them to use it. That’s a completely different thing than bolting on a brand-new system, and it’s worth watching for: the tool that’s already woven into how you work usually beats the shiny standalone one.
On early vs. late: this is where the adoption spectrum comes in (it’s in the toolkit). Her honest take: for most organizations that aren’t tech companies, it’s okay to be a later adopter, “because being at the front of the curve also opens you up to all those vulnerabilities.” So weigh the stakes — what do you actually have to gain from going early?
And if the answer is wait? Wait — but do it actively. “Wait and see is such a legitimate strategy. The main thing you need to do is actively do the see part.” Keep talking with peers. “You can ask a trusted advisor who’s deep in this to ‘keep an eye on this capability for me’” so that you’ve got eyes on the one thing you actually need, instead of drowning in everything that’s coming.
One last bit of order, because it matters: when you do go looking, Grace’s order of trust is existing partners and vendors first, then peers, then your own team — AND THEN your favorite reputable media source. Not the other way around. Start with the media and you just get swept into the hype and the fear.
→ Want someone deep in this to watch for the capability you actually need? That’s a great reason to keep Grace in your corner. Book 30 minutes with Grace.
When to do this yourself vs. bring Grace in
A lot of this you can do on your own — and you should. Mapping one process, layering on your tech, naming what stays human, putting a number on the friction: that’s all yours to do, and the toolkit walks you through it. Plenty of teams will get real clarity (and start a few simple automations) just from that.
Where it makes sense to bring Grace in:
- You’ve tried to map it, but thinking in workflow mode keeps tripping you up (it’s the hardest part — you’re not alone there).
- The friction is real but the fix is genuinely complex — like pulling data your software won’t give up, or stitching together a custom build.
- The stakes are high: money, privacy, or sensitive information where accuracy really matters.
- You’re staring down a build-vs-buy decision with real dollars on the line and want a straight answer before you commit.
- You want someone watching for the one capability you actually need, so you can stop tracking everything.
If you want help
Grace Kim is an independent data and AI consultant who helps mission-driven teams make the right technology decisions — starting from your real friction, not the firehose — and then actually put them to work in how your team operates. She does the process mapping, the automations, and the custom builds, and she keeps everything usable and grounded in the tools you already have.
She’s a good fit if:
- Feel the pressure to “do something with AI” and want a grounded, inside-out place to start
- Have a specific pain point and want to know whether it’s worth fixing — and what worth spending money on
- Are weighing build vs. buy and want someone who’ll tell you the truth
- Have sensitive data and need it handled with real care
Connect with Grace Kim!
What to take with you
Start here (free):
- Grace’s Tech Adoption Toolkit — map your process, layer on your tech, find the friction, and walk into any tech decision knowing exactly what you need.
- Find Grace and her work at mappinginsight.com.
Recommended reading:
- Grace’s newsletter, Human in the Loop — her perspectives and research notes on tech advances and what they mean for labor, the environment, and tech governance.
- Your Undivided Attention podcast by The Center for Humane Technology — how emerging technologies are reshaping society, and what it takes to steer them toward the best possible outcomes for humanity.
- IDEO’s “The Five Deadly Innovation Traps — and How to Avoid Them” — a framing Grace keeps in mind; a lot of her work is about avoiding exactly these.
Questions to Sit With
- What’s one process that’s been slow or quietly frustrating for so long you’ve stopped really noticing it?
- Where in your work does the human touch matter so much you’d protect it even if a tool could technically do the job?
- If you put a number — hours or dollars — on your most annoying inefficiency, what would it be? And what would that make worth spending?
- Who’s already in your corner that you could ask to watch for the one capability you actually need?
- How much do you want to steer the way your team uses AI versus letting people explore on their own? (There’s no one right answer here — it shifts with your size, your data, and what you’re working on. Worth revisiting on purpose rather than by default.)
- What’s a skill your team would quietly lose if you handed the whole task to AI — and is that a trade you actually want to make?
Not sure Grace Kim‘s the right fit? Talk to Helia directly!
This article comes from a coffee chat with Grace Kim in May 2026. These conversations form the heart of the Helia Library — because we’ve learned the most from doing and from talking with other doers willing to share their wisdom. We don’t need to start from blank pages or do everything alone.
As always, take what’s helpful, leave what’s not, and make it your own.
About Grace
Grace Kim is an independent data and AI consultant who came to this work the long way — through a PhD in earth science and years at NASA studying satellite data, asking big questions about how systems actually work. She left big consulting to make that kind of thinking available to organizations that could never afford a five-person tech team. (She also took our entire interview with a cat steadily clawing at a power cable in the background, which felt exactly right for someone whose whole philosophy is meeting the real world where it actually is.
Work with Grace
