Latest News An interview with Ahti Heinla of Starship Technologies

July 27th 2017

Starship Technologies is an Estonia-based robotic startup created in 2014 by two Skype co-founders: Ahti Heinla and Janus Friis.  Starship’s bots have been pegged by the NY Times as “a glimpse of the future”. Standing at two feet tall, weighing 40 pounds and travelling at just 4mph, Starship’s autonomous six-wheeled ground-delivery bots are set to revolutionise local delivery.

An interview with Ahti Heinla of Starship Technologies


A coding enthusiast since the age of 10, Ahti Heinla is Starship Technologies’ co-founder, CEO and CTO. Ahti was also a co-founder and Chief Technical Architect of Skype and KaZaA, plus eight other start-ups. In this interview we grilled asked Ahti about his plans for Starship and the future of last-mile delivery.

Is there a delivery paradigm shift on the horizon? Without exposing your secret sauce – what is Starship’s proposition to potential partners?

Yes, definitely. It’s undeniable that automated delivery methods are producing a paradigm shift as you say.

Not so surprisingly – consumers like it and businesses are reacting to it really well.

Our general value proposition can be boiled down to convenience and cost:

  • Cost is obviously a big driver in the delivery industry. It’s the lowering of operating and technology costs that has fuelled the rise of the last-mile delivery segment in particular.
  • Historically, it’s been really hard for large delivery companies to see the wood for the trees when it comes to optimising delivery costs as wages are increasing.
  • At the same time as wages are increasing, the cost of technology is going down, opening up a realm of opportunities for high-technology products to disrupt this industry.
  • We see there is no natural lower bound cost for robotic delivery. Ultimately the new introduction of robotic technologies such as the Starship bot and its ability to reduce the cost of deliveries to pennies on the pound means that this shift is inevitable.
  • Convenience plays a significant role. We’re now living in a world where consumer choice is king; people want their items really quickly and they want to know when they’re going to get them.
  • So the old customer journey model of “I know I’m going to get my delivery on Wednesday. But I don’t know when on Wednesday (and the delivery driver can’t leave the package on my doorstep). Do I really need to be home for the whole day to get my delivery?” is essentially dead.
  • Starship offers 15-minute delivery windows, issuing real-time delivery updates as standard. That’s something which would actually be very costly to match with traditional delivery methods.

Eventually I can see our product evolving to suit various delivery models, in the short term it’s quite clear that we can cover most of the needs with just one model.

It just makes it a lot simpler if you have one model, especially if you’re conducting smaller pilots. It makes sense to look at different models when we’re at the stage of manufacturing quantities of 100s of thousands.

What competitive advantage does Starship provide over other delivery models? And, what does your short-term growth strategy look like as a result of your recent £16m investment round?

Other delivery models will still have a place in the future – so it’s not that the Starship approach is going to replace everything. Different delivery methods are suitably different.

Bike couriers are certainly optimal in very dense urban environments as they can overcome gridlocks and traffic jams etc. but if you’re talking about the suburbs for example, when the traffic is comparatively low and it’s relatively easy for an autonomous vehicle to operate there, then I certainly think that automated robots are better placed for the job.

As for drones, I generally think that people don’t like drones flying over their backyards with other people’s parcels dangling from them, and they’re also susceptible to operating cost issues.

In general, flying machines are difficult to make – they need to be ultra-reliable because there are so many different safety concerns, and aviation regulations are difficult ones to overcome.

We’re steering around all of these issues by staying on the ground, this also allows us to use simpler technologies.

Will Starship bots make it onto the roads?

It’s a timing issue. So, in the long run, yes!

But I don’t really believe that autonomous vehicles in a generic unattended urban environment (so, Level 5 autonomy) will actually make deliveries any time soon.

OK, maybe a little bit in isolated trials but I don’t think that this will be something that will happen next year. Whereas, with the Starship model, our bots will actually be operating next year.

So, I predict that it’s going to be a while before roadworthy autonomous delivery vehicles are prevalent on UK roads, and I think this view is supported by the majority of industry predictions.

So, in general we’re still just in testing mode right now. Starship is obviously expanding, so we’re getting robots into service and building all of the infrastructure around them ready for scaling with the aim that next year is definitely the year where you’ll start to see our robots appear in large numbers.

We’re hiring a lot of people as a result of our latest investment round and talent is definitely a key driver for success for any company at this stage.

Are you positioning Starship to own the instant delivery category? Or, are you working with partners to disrupt other areas of last-mile delivery?

We want to take care of the instant delivery segment of the last mile, but obviously we need to work with partners to do this. Our commercial partners are usually producing or selling the things that we are delivering.

Like we touched on before, instant delivery is being driven by rising consumer expectations so a lot of businesses are looking to offer this level of service to create competitive advantage.

So it’s a combination, other areas of last mile delivery are also very appealing – but one thing is for sure, we cannot work in complete isolation.

We’re in contact with a lot of people in the industry, more than a 1000 companies have contacted us with the intentions of partnering. They all operate with various business models so we’re exploring many different things – which is really exciting to be part of.

If this is a market segment you’re looking to disrupt, how will you overcome drop-density related challenges?

Drop density – along with other factors – form part of an equation which determines whether certain modes of transport, logistical schemes or models are viable or not.

Drop density is actually something that’s not very easy to control. It’s outside of our control for sure, but we need to make our service work with the greatest range of drop densities that we can accommodate.

A lot of that is actually down to operational efficiency and operational costs – and as I said, robots and technology, in general, have no natural lower bound of cost.

So, suppose it will at some point cost 5p per mile to operate an autonomous robot on the sidewalk -then at that point, drop density doesn’t matter, you can deploy a robot to serve just one customer 5 miles away regardless of how many other customers there are nearby.

The logistical models that can be used with robots are different than those models satisfied by traditional delivery methods. All of the concepts around “hub and spoke”, parcel containers, robotic warehouses and others lend themselves really well to robotic delivery.

It’s a match made in heaven. At Starship, we have an ongoing collaboration with Daimler who is obviously a big manufacturer of delivery vans. Together we’re building what we call a “RoboVan”, which is a range extender for the robots – this is a great example of a mobile hub or parcel container.

So all of those models are possible and we’re definitely hard at work looking at them, simulating different models and finding out what works with what drop densities and different kinds of customers.

Are there any other product categories, beyond food delivery, that lend themselves to this form of delivery? For example, pharmaceuticals?

Pretty much everything actually. So, food delivery is an area where the instant delivery model is inherently required as people want to eat now. But actually, when given a choice, people want everything else instantly as well.

So, if they’re ordering a book from somewhere they also ideally want to get started reading it immediately.

You’d be surprised how spontaneous people are when buying products online – it’s usually because they have an appetite to consume that product now whether it be books, electronics or food.

So eventually it will become commonplace that instant delivery will be the norm for pretty much everything. People are coming to expect it. And for us at Starship, actually food delivery – isn’t a disproportionate focus of ours.

We are exploring a lot of different sectors, and as a result we have a number of exciting partnerships (that we cannot talk about yet, unfortunately) where we’re working with a number of different companies across a whole host of different sectors.

A good example of an industry that illustrates the diversity of our model is the flower industry.

Not many people have thought about how robots lend themselves to delivering flowers, but actually it works really well. We did our first flower delivery in the US about a month ago and it’s been working great ever since!

Like I said we’re working with many different companies. I don’t necessarily think that in 5 years’ time people will really divide things into “this is an instant delivery category product, and this is a non-instant delivery category product”, I think when instant delivery becomes more available and affordable through the use of robots and automation, people will want it for every product that they buy.

What things are you looking at in particular to reduce the cost of the Starship product? Can Starship’s operating model be easily transferred from the sidewalk and onto regular streets when regulation and level 5 autonomy permits?

We’re definitely doing a fair bit of cost engineering right now, and this takes many forms.

Traditional hardware engineering is one key area that we’re paying particular attention to in order to make sure that the robots are cheaper to produce and require less maintenance.

So this is definitely something that we’re working on. Operational costs are a key driver for change, so we’re always looking to refine various operational components.

For example – we’re building a hybrid autonomous robot which can drive autonomously (the majority of the time) and then be teleoperated by a human operator sitting in a call centre type setup – via live video link.

Reducing the percentage of time that the robot needs this kind of help is an important mitigating factor in minimising our operational costs.

Plus, a lot of optimisation has all sorts to do with better routing and the different logistical schemes – for example when I described the RoboVan earlier, actually the point where the RoboVan dispatches robots and continues on with its long journey, we have a choice to take: we can use robots for longer distances or for shorter ones and this actually matters with the overall cost equation.

So we’re doing some quite significant work in fine tuning those scenarios in order to optimise cost.

I don’t necessarily think that road-going autonomous vehicles would be so much better than pavement robots in delivering.

Yes, sure, they’re faster. So let’s say they’re on average 5 times faster than the other robots, but then some things inevitably offset that quite a bit – e.g. when the vehicle gets to the consumer location there still has to be the handover of the package and this takes a certain amount of minutes, and that part is not faster in the larger road-going autonomous vehicles.

So, the overall time efficiency that the faster driving vehicle would get is actually not that great, but at the same time, it would be a lot more expensive, it would have a higher insurance cost and the operational cost efficiencies would start to become imbalanced and unpick the model if you looked at it closely enough.

So, I anticipate that for at least the next ten years’ efficiency actually works in favour of pavement delivery. And we can do it now! Big businesses can use our technology now to streamline their delivery operations! And that’s the point that underpins all of this.

How far away are you from level 5 sidewalk autonomy?

We’re practically there. So we can drive without being supervised at certain places, and we’ve done so for a while now in various locations across the world.

And we’re going to expand this service further. When I said that “towards the end of this year we will look to scale our operation”, I definitely don’t mean that we’ll scale with tonnes of people following the robots and supervising them all the time. We are fairly close, you could say it’s happening right now.

Have you developed all of this in-house? No-one has attempted this in the past so everything you’re doing is first of kind in nature.

Absolutely, we’re doing a lot of things on our own. In fact, some people think that we’re doing too much on our own! But I think the strategy we’ve chosen in our product development is right.

Do you think the transparency delivered by Starship Droids (i.e. the ability for consumers to know, understand and anticipate their delivery through real-time feedback) outweighs the need to have a face-to-face interaction between the consumer and the delivery driver?

Yes, we certainly think that. And we can compare this to how booking flight tickets worked 20 years ago.

You know 20 years ago it was commonplace to book your tickets via a travel agent. People called up a travel agent or visited them in store to talk to a human about where they wanted to fly to and when and so forth.

Travel agents then presented customers with the options and prices etc. and you made a decision.

So all of that has evolved since the dawn of the internet and people prefer it. Of course there is some human interaction that is lost in the process, but at the same time, people certainly appreciate that they can book their flight tickets whilst they’re in the comfort of their own home at any time of the day.

People appreciate that level of convenience and transparency and in fact I don’t really remember the last time I heard anybody saying that they missed not having to call up the travel agent. I don’t think I’ve ever heard anybody saying that. And I suspect that the same will happen with online delivery.

Is there anything else preventing the industry from turning this vision into a reality (path-planning, interactive navigation, autonomous localisation in crowded environments, software/hardware)? What timeframes are we working towards?

There is no big barrier or big obstacle ahead. All of those things that you mentioned just need mastering. For a lot of those things, it’s not enough to have something at a prototype level for commercial delivery, they need to be fairly well defined.

And we’re walking that path; developing our own technology and business model and refining it, all to get to a point where it’s all about scaling.

We’re working fast and I don’t really see any big road blocks in our way, of course there’s work that needs to be done, but it’s happening because we’re doing it right now.

Has the UK’s ability to tweak their regulations, their increasing labour costs, their relatively pro-innovation approach to entrepreneurialism and the general public’s positive sentiment towards robotics been a driver to adopting HQ in London?

Absolutely. Plus, my co-founder Janus Friis who was also a co-founder of Skype lives in London, but in general, London is a global city and the focus of a lot of innovation and delivery activity.

The adaptable regulations and the overall level of innovation in public administration are certainly high and we’re also in contact with authorities in the UK who have a great attitude towards innovation and automated vehicles.

We haven’t really tapped into any UK networks for support, we are working very independently and haven’t collaborated very much with existing networks.

That’s not to say that we won’t look to do this in the future but it will be for technical guidance opposed to funding as we’re funded by the venture funding community.


If you enjoyed reading this, you might enjoy reading: 

Made by Union Room