David Kennedy

David Kennedy spent 26 years at Aer Lingus, including more than fourteen years as chief executive of the airline.

In the early stages of this career David focused on information technology and led Ireland’s most ambitious software implementation project. This involved installing the largest mainframe computers in the country, building an international data network, growing an in-house data processing department and creating the Cara subsidiary to sell computing services to other organisations.

My own involvement in Aer Lingus started in 1962 when I replied to an advertisement for an operations research (OR) analyst. I was recruited to work in a newly created operations research unit under Finbar Donovan, who was systems manager with responsibility for data processing, method study and OR. I had just completed a masters degree in experimental physics in UCD.

Finbar was the person who pioneered computers in Aer Lingus and could reasonably claim to have been the leader in commercial computer applications in Ireland. He also came from a background in physics as a graduate of UCC. He joined Aer Lingus in the late 1950s with a mandate to introduce the most modern management skills with an emphasis on the application of developing technology. Such an appointment was very much in the style of JF Dempsey, the then general manager. JF had a strong view that Irish people were well capable of running businesses with high technological requirements, a view not widely held at that time.

My first assignments included the development of mathematical models for traffic forecasting, using exponential smoothing and the use of queuing theory to estimate the staffing needs at different times of the day in telephone sales and booking offices. I was introduced to John Byrne in TCD and he assisted in writing a computer program for traffic forecasting on the university’s IBM 1620 using the Fortran language. My recollection is that the entire CPU had a core memory of 64k, which meant that one of the demands on the programmer was to be very skilful in the use of memory. Although TCD was the first third level to have a computer department, they were followed soon afterwards by UCD where Frank Anderson followed suit.

In 1964 Finbar Donovan was appointed sales manager for Aer Lingus. This meant that, although he no longer ran the data processing department, he retained responsibility for reservations which was part of the sales department. The data processing department included telecommunications and a fairly basic IBM 1440, used primarily for revenue accounting purposes with punched card inputs.

Finbar stayed in touch with IBM who were working closely with American Airlines on the development of a real time computer system for airline reservations. In anticipation that Aer Lingus would intend to move in this direction, an internal team of about eight trainee programmers was recruited internally and provided with basic training by IBM. Members of that initial team included Tim Lane, Emmet Wilson, John Blunden, Brian Ennis and Eamonn Giblin. Planning proceeded in the sales department with IBM through 1964 and 1965 for the possible advent of an IBM reservations system based on the IBM 7010.

However the joint IBM/American Airlines system failed to deliver as the 7010 was not powerful enough to handle such a complex real time system. Finbar and IBM together worked out an interim system using a Bunker-Ramo Teleregister computer, linked by telecommunications lines to the main reservations offices, which showed whether flights were open for sale or not. It had its own skilled maintenance team with people such as Michael Rogers who were also responsible for all other telecommunications and telephones throughout the airline. At the same time another IBM 1440 was put in place which updated reservations bookings through punched cards although not on a real time basis. It filled the gap but fell well short of a full online real time system.

Senior management in Aer Lingus became unhappy about the fact that there were now essentially two separate computer departments in the airline. I was asked in 1965 to write a confidential report outlining my views on how to proceed against the background of computer developments in other commercial organisations. I surveyed what was happening elsewhere in Ireland (not a lot with some exceptions such as the Sugar Company) and the UK. My main recommendations were that future IT investment needed to be subjected to more rigorous cost benefit advance analysis than had previously been the case and also that it made no sense to have two separate computer departments in Aer Lingus. This obviously resonated as the company implemented my recommendations.

Toward the end of 1965 there was a significant management shuffle in Aer Lingus arising from the imminent retirement of JF Dempsey (to be succeeded by Michael Dargan). Finbar was appointed assistant general manager, administration and I became systems manager, reporting to him and responsible for a newly united systems department which now included all IT activities. Shortly afterwards Finbar left Aer Lingus to work with AIB and subsequently with the development of a travel agent based reservations system. I then reported to GP Dempsey who was the chief financial officer.

One of the Boeing 707 aircraft that Aer Lingus operated in the late 1960s.
Photograph source: Nora O’Rourke

In the meantime IBM announced the development of the System/360 series and indicated that they were now in a position to restart work on a real time reservation system known as PARS in collaboration with American Airlines. This was probably the first time that a real time system would be applied to commercial activities and it entailed a quantum leap in the level of complexity compared with the traditional batch processing computer applications. Shortly afterwards IBM announced that a version of this system called IPARS would be developed for international airlines, where the reservations requirements were somewhat more complicated than for US domestic carriers.

This was of immediate interest to Aer Lingus and intensive discussions took place with IBM leading to a formal offer whereby IBM would sell us all the necessary hardware and make available the IPARS software free of charge. The price quoted was just under £2 million which represented a very large investment at the time. In order to gain confidence in this proposition we had discussions with American Airlines IT people who were very open to sharing their assessment of risk. I also met with Univac and Burroughs who were working on a similar development. However it was clear that although some of the other US carriers and some European carriers were working closely with Univac in particular, nevertheless IBM had a clear lead in the race. IBM signed contracts with Pan American and Eastern Airlines, in addition to American, with the intention of going live with PARS in early 1968.

In January 1969 a group of students from the UCD Mathematical Society visited the Aer Lingus computer centre for a demonstration of the airline’s new reservations system.
In this photograph David Kennedy (on the right) demonstrates one of the two IBM System/360 model 50s at Aer Lingus to Murt O Sé, Cathal Brugha, Eoin Bairéad, Paul O’Neill, Denis O’Hogan and Martin Walsh. Jean Mallon, who worked in the systems department, is seated at the computer.
Photograph source: Eoin Bairéad (Originally published in The Irish Press)

The issue for Aer Lingus was whether this project represented an excessive level of risk and whether it might be better to defer consideration until IPARS had been exhaustively proven elsewhere. The proposed timetable for us that we would follow immediately after BOAC in summer 1968 and ahead of KLM, Swissair and Alitalia among the European carriers towards the end of 1968. In evaluating the proposal the immediate benefits of a real time system were obvious and to a significant extent measurable. The direct costs were also relatively easy to quantify. The downside lay in the risk that the system might not perform as specified and in particular that there would be long periods of downtime during which all flight bookings could not function with serious implications for passenger service and revenue generation. My considered view, based on detailed discussions with IBM, with colleagues in other airlines and our own professionals, was that this risk was limited and manageable. Incidentally a comparable approach was adopted in selecting aircraft, when Aer Lingus around the same time became a launch customer for the newly developed Boeing 737 aircraft.

There was considerable debate at senior management level in Aer Lingus with the senior finance management strongly opposed to the project as being too risky. With senior management support elsewhere Michael Dargan encouraged us to proceed. This led to intensive negotiations with IBM where Aer Lingus pressed for some degree of risk sharing on the project and with IBM reluctant to do so. Eventually a level of compromise was reached whereby IBM in a side letter agreed to use “best endeavours” in the event of serious problems arising.

One issue was whether the System/360 CPU should be a Model 50 or the more powerful Model 65. The latter would have added significantly to the cost but was the choice of the other airlines planning for IPARS. IBM encouraged us to consider the Model 365, but we were able to establish on the basis of detailed analysis of peak reservations message from all the telephone sales offices throughout the airline that the Model 50 would be more than capable of handling peak traffic loads. The concept was to have two mainframe computers, one working full time on IPARS and the other acting as a back up and also capable of handling all the other batch processing work in the company. In view of the crucial nature of reservations in an airline, a separate back up facility was planned with IBM where all daily activities were recorded. In addition an emergency power supply was installed.

It was clear that considerable preparations were needed in order to cope with the advent of what I termed Automated System of Telecommunications and Reservations for Aer Lingus (ASTRAL). A number of working groups were set up as follows:

1. A software team under Dennis Behan would be the main link with IBM. The team included a number of systems programmers drawn from the early programmer team and two very experienced programming specialists, one from the US and one from the UK, who were hired on contract.

2. A telecommunications team under the direction of Pat McCarthy, an experienced engineer in this area, who was recruited from RTE and who took over the former Bunker-Ramo team.

3. A computer operations team to run the operation on a 24 basis, while continuing to run the punched card division which was still needed for the increasing requirement for batch processing elsewhere in the company. This was the responsibility of Liam O’Cassaide and Liam Grimes.

4. A reservations control section, run by Tom Clerkin in liaison with the programming team, to define the specifications for software changes to IPARS that Aer Lingus required.

5. A training group under Johnny Russell, who was responsible for booking offices and telephone sales in Ireland, to ensure that all the sales staff were trained to handle the new systems. This included staff not only in Dublin but also Cork, Limerick, London, Birmingham, Manchester and Glasgow, all of whom were going to go online at the outset. Close to 200 people had to receive training in the operation of ASTRAL.

6. A number of separate programming teams to continue the development of batch processing projects elsewhere in the company, mainly in the finance group but also in inventory control of aircraft spare parts (and subsequently elsewhere). John Quill was one of the senior systems analysts in this area.

In addition to the above personnel Paddy Byrne played an important role. He had been in charge of the early IBM 1440 development but had since moved to become assistant cargo manager. At my request he came back to act as assistant systems manager to support the project. Paddy later became systems manager when I moved elsewhere in the company.

The main feature of the new system was that it controlled all reservations made on Aer Lingus flights and was directly accessible in real time by all the telephone sales agents and booking offices which were directly linked to the CPU through high speed phone lines. Booking requests for flight segments on other airlines were processed by automatically generated telex request messages. Requests for hotel bookings and car hire rentals as part of an airline booking were processed in similar fashion.

Most of the IBM hardware was delivered in early 1968. This comprised the two CPUs, tape drives, disk drives and modems. The IPARS software was released gradually in the coming months. IBM had an expert team testing it out with the Aer Lingus team and supporting the changes needed by Aer Lingus. The support from IBM under Dick Cahill was excellent and there were only a small number of hitches as the software was introduced. During the early summer of 1968 we paid a number of visits to BOAC who were very open in sharing their experience.

ASTRAL finally went live in late 1968 without any major fanfare and with surprisingly few problems. All the airline’s sales offices in Ireland and the UK went online from the beginning. The following year the office in New York was linked up using a dedicated high speed line via transatlantic cable. On that occasion the Minister of Transport and Power visited Dublin Airport for a demonstration of the new technology and was suitably impressed to be able to view a booking request made in the USA being handled in Dublin. As communications costs began to fall most of the sales offices in the other major European cities were gradually added.

A subsequent cost benefit analysis of the system showed an excellent return on investment and a payback period of between three and four years. There were significant staff savings in reservations control and savings resulting from the reduced time required for agents to handle telephone bookings. These alone would have paid for the development before taking into account higher load factors on flights which, although real, were difficult to quantify.

Once the project had been successfully completed it became clear that Aer Lingus had a level of expertise and also surplus computer power above its immediate needs on the back up System/360. With the encouragement of Michael Dargan we launched a systems services division under Maurice Foley who had worked alongside me as an operations research analyst. The objective was to create a profit centre from our existing resources, both hardware and software expertise. Our first customer in 1969 was An Bord Bainne and a number of others followed shortly afterwards. Subsequently Aer Lingus set up Cara Computing as a separate corporate entity and later Dennis Behan became its chief executive.

Some years later ASTRAL was extended to incorporate passenger check-in at airports. In addition Aer Lingus staff supplied their reservations expertise on contract to a number of other airlines in the early 1970s including taking full responsibility for the launch of the reservations system in Kuwait Airlines.

In 1969 I became general services manager with responsibility for other departments, such as general administration and airline catering, added to the systems department. In the following year I was asked to take on project management for the construction of the first Aer Lingus hotel, the London Tara and in 1971 I was sent to New York as senior vice president for Aer Lingus in North America. After two years there I returned to Dublin, firstly as assistant chief executive operations and then as chief executive in early 1974.

Last edit: June 2019

© David Kennedy 2019